Intelligent Machines 886 transcript
Please be advised that this transcript is AI-generated and may not be word-for-word. Time codes refer to the approximate times in the ad-free version of the show.
Leo Laporte [00:00:00]:
It's time for Intelligent Machines. Jeff Jarvis is here, Paris Martineau, our guest Div Garg, young guy, about 27. He thinks the future of AI is not big models, but little models that do lots of little things, including your phone, including ordering your groceries. Div Garg, our guest, all the AI news too, and a big announcement from Micah Sargent coming up next on Intelligent Machines. Podcasts you love. From people you trust.
Paris Martineau [00:00:29]:
This is 20/20.
Leo Laporte [00:00:35]:
This is Intelligent Machines with Jeff Jarvis and Parris Martineau, episode 886, recorded September 2nd, 2026. Infinite slop. It's time for Intelligent Machines, the show where we cover the latest in AI, robotics, and the smart devices all around us, including those who have no power. I am in a blackout right now, but fortunately I have battery backup. And fortunately, We have backup in the studio with us. Before we do that, though, let me say hello to Paris Martineau from Consumer Reports, our intrepid reporter on all things poisonous. Hello.
Paris Martineau [00:01:12]:
You know, they finally found a way to cut you off from all of your AI agents.
Leo Laporte [00:01:16]:
They did. The agents have died.
Div Garg [00:01:19]:
Oh, no!
Leo Laporte [00:01:20]:
No, they're still alive.
Paris Martineau [00:01:22]:
Leo, at some point, might just disappear if he doesn't get power back.
Leo Laporte [00:01:27]:
The battery's running out.
Paris Martineau [00:01:28]:
And Pam and I initially were like, he's just wandering around shouting Quicksilver in the dark.
Leo Laporte [00:01:34]:
No, they're alive still, but, but I don't know how much longer. They're both— everything's backed up on a UPS.
Paris Martineau [00:01:40]:
Are the agents draining the battery too?
Leo Laporte [00:01:42]:
Is that what they were holding on to? Well, they have their own battery. Everything— everybody's— we're all sharing right now. Anyway, yeah, that could happen. As you can see, there's nothing behind me because I did not put the neon sign on battery backup, but my studio lights are. Anyway, hi Paris, also with us. And unfortunately, the blackout is not going to affect the jingle. It is the Emeritus Professor of Journalistic Innovation at the Craig Newmark Graduate School of Journalism.
Paris Martineau [00:02:12]:
The Craig Newmark jingle is on battery power.
Leo Laporte [00:02:15]:
Definitely battery backed up. Author of Hot Type, which is, I think, probably selling like hotcakes.
Jeff Jarvis [00:02:22]:
We hope.
Leo Laporte [00:02:24]:
Do you get— do you get like—
Jeff Jarvis [00:02:25]:
I don't even want to know.
Leo Laporte [00:02:26]:
Your agent doesn't call and say, Congratulations, Jeff, you're number 1 on the charts.
Paris Martineau [00:02:31]:
I gotta buy more copies. We gotta get you a little dagger next to it. Me and Leo will buy 1,000 copies each.
Leo Laporte [00:02:38]:
We have a very interesting guest we actually prerecorded an interview with. His name is Div Garg. He is the founder and CEO of a company called AGI, the AGI Company. And Div's background is pretty amazing. He dropped out of a PhD program at Cornell to work on AI. He has worked pretty much everywhere, but you know what? Instead of me telling you, why don't we ask Div Garg, our guest on Intelligent Machines? I asked him, how did you get all this going here? How'd you start? Div, so tell me, so you have a really interesting career, first of all. Undergraduate at Cornell where you invented pseudo-LIDAR, Right, which is basically LiDAR from a camera.
Div Garg [00:03:30]:
Yes.
Leo Laporte [00:03:31]:
Am I right? Yeah. Worked at NVIDIA. Of course, first thing I would do if I were NVIDIA, Jensen Huang said, come here, come work for us. Did you— you stopped— you dropped out of your PhD program to do that, right?
Div Garg [00:03:42]:
I did, yes.
Paris Martineau [00:03:43]:
Yeah.
Leo Laporte [00:03:44]:
Apple Special Projects under Ian Goodfellow. That's a pretty nice credit. Tell us the truth, were you working on the car? Uh, no, you weren't. You couldn't say if you were.
Div Garg [00:03:58]:
The limits, please.
Jeff Jarvis [00:04:00]:
That's all right.
Leo Laporte [00:04:01]:
They're just to kill us now. Shred us. Um, I can go on. I mean, just all sorts of interesting stuff. But the latest is AGI Inc. Tell us what AGI Inc. is.
Div Garg [00:04:15]:
Yeah, like, uh, so, so one thing we're trying to do is, uh, Like, how do we build the next generation of AI interfaces? So even right now, if I'm using AI, like, I have to go to a laptop and like maybe like use my Claude or Codex or whatever like you like to use, right? And our belief is like, it's like we really want to have this next generation of AI-native devices. But if I'm using my phone, my phone should be AI-native. Like, you should have something like Siri that actually works and does things for you. So like, how can we make every piece of device you own, uh, have an AI built in, and it can actually go and like do anything on your device for you. And, and then like, that would be wonderful if I can just like talk to my phone, it like basically like books me an Uber, gets things done, uh, sends my emails. We really want to go and live in this world where we remove the need for like typing and like spending all the time on a screen. And that's where we're like, like really, really going to do like, how do we build this new interface. And like, how do we like champion this effort? We're like, okay, now you can spend your time actually focusing on your life and getting things done and not on your screen.
Paris Martineau [00:05:22]:
So you decided to name the company AGI Inc. A game we always like to bat around here is what is your definition of AGI, and has that changed at all over the past couple of years?
Div Garg [00:05:34]:
Yeah, I feel like it's like a, um, like an ever-growing definition. I feel like the best way to say is like You have reached an AI capability where, like, whatever, like, agent or intelligence you have is smarter than an average human. And I think that's probably the bar. So, like, maybe, like, whatever an average human can do, if the agent at some point can go do tasks at a slightly better capability at, like, pretty much everything that you can throw at it, I feel like that's basically a good definition of AGI to start with, because then you can say, like, this is already better than, like, a human brain. And then like, uh, and then it can create like economic value for you. Um, so that's how we measure it. Like, we're like, can it be— can it actually do things better than like an average human?
Leo Laporte [00:06:18]:
Well, I mean, a calculator can do arithmetic better than an average human, but I wouldn't call it AGI. Uh, AI is spiky, right? So I think right now AI can code better than the average human, maybe not the best coders, but better than the average human.
Jeff Jarvis [00:06:34]:
Can it code better than you, Div?
Div Garg [00:06:37]:
Uh, maybe at this point, who knows?
Leo Laporte [00:06:40]:
We know as of the Robot Olympics, they can run faster than the average human, but so can a car. You don't call a car intelligent. And I think there are plenty of things— you, you must know this from your own research with self-driving vehicles— there's a lot of things that are really hard at the edge, like that last 5% of self-driving. That yet AI cannot do effectively. I wouldn't, I wouldn't drive off and just go to sleep and let my Tesla take me home.
Paris Martineau [00:07:08]:
So some people do though.
Jeff Jarvis [00:07:10]:
That's true. The next question is, uh, once you've made your definition, uh, how far away are we from that state?
Div Garg [00:07:17]:
Hmm. So like one thing I like to do is like separate like, uh, something that's like a digital AGI versus a physical AGI in a sense. Um, and that's basically like, okay, anything that I can do in the digital world, like on a computer or like—
Leo Laporte [00:07:31]:
Ah, that's good. Yeah.
Div Garg [00:07:32]:
And that we feel like, it's like, okay, that's getting there. Like anything that a human can cognitively do, and here I'm considering all dimensions, so not just like one spiky dimension, like anything that a human can do that's a cognitive digital task, if AI can do that better, I would say like that's like digital AGI in a sense. And then you have to like kind of like be very all-encompassing, think about all the dimensions. And I think that we might have that by end of this year to some extent. Like it's like, I think like the technology is getting really good.
Leo Laporte [00:07:59]:
We're very close to that. I mean, and in a way, it's the AI's— it's a computer talking to a computer. It's its native tongue, more so than it is a human native tongue. I think it's really interesting that you're focusing, though, on user interface. That is always historically, you know, I remember when I first started using computers, the dream was to be able to talk to it. In Star Trek, hello, computer, use the mouse. That was always the dream. And we're getting— Pretty close now to being able to interact with your computer in, in plain English.
Div Garg [00:08:34]:
Yep. Yeah, even when Alan Turing like first came with the concept of, of a programmable machine, right, like he was thinking about it like he was inspired by humans. He was like, he wanted to go like, if you have seen like the, his biography or the movie The Imitation Game, he was inspired by his best friend. Like he wanted to like kind of build something that was like a human he could talk to, has intelligence. So computers in a sense like came out of human desire to build something that's an artificial human that you can talk to. And I think maybe we spent like almost like maybe like a couple of decades, around 50 years or slightly more, where we are now reaching to the initial dream of why computers came to be.
Jeff Jarvis [00:09:13]:
So do you see your company producing more for consumers or companies? A, and then B, is it, is it a platform where we can all make our own agents?
Div Garg [00:09:26]:
Yeah, that's a good question. I also think we like to think of it like this, like we're basically building this agent as a capability. Like, so like how maybe you might be using Claude, right? So like Claude is something that people use daily, like a lot of businesses use it, but a lot of individuals also use it. Same thing for like maybe like other kind of agents, like OpenCLaw, for example. So I feel like it's very hard to divide agents into like, is it B2B or is it like a business use case or is it consumer use case? Because like, what happens, like, these are kind of like almost a general platform where you just throw anything, it gets things done. So you're focusing on capability. Okay, like, can I get things done, uh, on the device by just like talking to it? And so it's like, you just like give it like a voice intent and things get done. You don't have to like spend your time like thinking about it.
Div Garg [00:10:10]:
And like, that's how we think about it, uh, on what we're doing. And this could have a lot of business use cases, this can have a lot of like consumer use cases, um, and then we have different ICPs, different One of the keys, of course, is getting the model on these devices, I think.
Leo Laporte [00:10:26]:
I mean, if you're going to have intelligence at the edge, you can't have it be phoning home to a giant server data center in Iowa. You have a very small model, AGI Zero, that is 2 billion to 20 billion parameters. So it would fit on a phone.
Div Garg [00:10:45]:
Yes.
Leo Laporte [00:10:45]:
But now my experience at this point, I mean, if you say that Fable is 10 trillion parameters, obviously that's It's not going to fit on a phone. It's not going to fit on a computer. The smaller the model, generally the dumber the model. Can you get this kind of order of intelligence on a model that small?
Div Garg [00:11:04]:
You actually can. I think the trick is you have to specialize. So what's happening with something like Babel is they're trying to throw every single problem possible in one model. And it's trying to do coding, it's trying to do math, it's trying to write poetry, trying to do anything possible. But if you actually spatialize, we found it's like even if you have a 20B model that's spatialized to one task. So suppose a 20 billion model specialized for coding can actually be really, really good and maybe reach close to Opus-level coding performance if it's trained specifically for just coding. And then you can do the same thing for mathematics. You can do the same thing for the kind of things we're doing with computer use.
Div Garg [00:11:42]:
So the trick kind of becomes you want to start building this kind of spatialization. And that's actually how the human economy works. It's like you don't have people who are trying to do everything. That's economics 101. You want to specialize. And I feel like now we're starting to apply that lesson to AI. You just have to have a specialized team of models. And these models can then become specialized and small enough that they can reach much, much higher levels of capability for their sizes.
Leo Laporte [00:12:09]:
So are your models dense models, the MOE models? Do you shard them? Bit by bit? How do you do this?
Div Garg [00:12:17]:
Yeah, we do a lot of stuff with MOE because that helps us, uh, be very memory efficient. And, uh, we're starting to build, um, a lot of stuff around like domain adaptation. So a core piece of our technology is that it can control any interface. So it's like it's gonna click and type for you, whether that's on like a, like a phone screen, uh, like this one.
Leo Laporte [00:12:40]:
So you'd have, you'd have vision and you'd have some sort of capability of interacting with the screen, basically.
Jeff Jarvis [00:12:46]:
Yeah.
Div Garg [00:12:46]:
And what we're deciding to do is like, uh, MO is great for spatialization. So we have like some categories of apps, might be shopping apps, or you might have like browsing, or you might have like something like— and then you can like activate the right layers depending on the domain.
Leo Laporte [00:13:00]:
That's how you get a 20B model on a 16GB iPhone, I guess. Yeah, yeah.
Paris Martineau [00:13:08]:
Uh, so walk us through a little bit— go ahead— the challenges and I guess the process for building out a model as well Because it seems like a lot of your recent announcements have been around kind of an MCP for all the apps on your phone. And you kind of famously earlier this year proclaimed that like the apps are all dead. What are the challenges? And I guess walk us a bit through the process of getting an agent to be able to seamlessly navigate a bunch of different interfaces on mobile devices specifically.
Div Garg [00:13:40]:
Yeah.
Div Garg [00:13:40]:
So as I said, there's a lot. Like the closest analogy is like trying to build a self-driving car. for your phone, in a sense. Because like, it's like, it's like, imagine there's millions of apps out there. Like, if you go to Google Play Store or the Apple App Store, there's like so many different things, and people have great interfaces and UIs and layouts. So you're trying to build something that can navigate every single digital road, in a sense. And there's like so much variety. And the challenge becomes like, how do you scale this thing? So like, maybe like you can build something that can work on the top 20 apps, but then how do you like work on the top 100 apps? How do you work on the top like 10,000 apps? And that's how you climb the frontier.
Div Garg [00:14:13]:
And I feel like even if you look at the industry, like, I think people have been working on computer use, browser use for almost 3 years now. So like, for example, like, uh, my last company was actually specializing in like a lot of browser control. OpenAI has been working on 3 years on like browsers, but I don't think they have still figured it out. So there's a lot of challenge on like, like solving to the, the, it's like there's like a long tail and you can get to 95%, but then you have to go to 99%, then you have to go to like 99.999%. And that's kind of the challenges we're solving. Like, how do we make these models continuously learn and improve over anything that you can throw at it?
Leo Laporte [00:14:47]:
That's a big challenge. Models that can learn and have memory and can get better. Are you doing anything that's unusual here, or is it just a bunch of memory.md files? How are you doing that?
Div Garg [00:15:03]:
Yeah, we actually have a great team of, I would call it like reinforcement learning experts. So we have a lot of PhDs.
Leo Laporte [00:15:09]:
That's your specialty, right, is reinforcement learning? In fact, I saw in your bio you taught a class in Transformers at Stanford.
Div Garg [00:15:18]:
Yes.
Leo Laporte [00:15:19]:
Yeah. So, so, so RL is kind of the key here.
Div Garg [00:15:24]:
Yes. And we have this trick where we let the agent autonomously go and learn on new apps. So we call this like self-play. So like the agent can go and like maybe can download a Japanese app. And then it can just go and run a lot of iterations. It will just try to map the UI, try a lot of different tasks, and then it will learn from its own experience. It's similar to how maybe we as humans learn how to ride a bike. You fall a lot initially, but eventually you get better and better, and then you become an expert.
Div Garg [00:15:57]:
It's kind of similar. The agent can self-play and it can learn the skill to now navigate any new app. And we are running a lot of these experiments, which allows it to like become an expert on new interfaces and new applications.
Leo Laporte [00:16:11]:
Does AGI Zero do that with its vision and action model? Does it— is it able to learn a new app?
Div Garg [00:16:17]:
Yeah, we actually have that capability now. We are— we've been like selective on rolling it out. So we have some partners we are giving this access to, along with spatialization, because it's like we don't want— we can't do everything. But then if we like choose like, here's this app we want to learn, and then we can learn that in like less than a week.
Leo Laporte [00:16:34]:
So, so the— eventually you don't need apps at all. Uh, you're— what you need is an interface to a local AI that can do everything an app would do for you.
Div Garg [00:16:46]:
Yeah, and I think, I think that's the vision. Because again, if you— if I walk back before phones existed, right, uh, I don't think people were like, like, I, I want to go and use an app for every single need I have, right? So like, if you think about it, like, apps are basically an abstraction of human needs. Like, okay, I want food, There's an app for that. I want to get a ride. There's an app for that. I want to book a plane. There's an app for that. So what happened is all our human needs got codified digitally into an app.
Div Garg [00:17:10]:
But I don't think people, when they started, they were like, okay, I want to have this ecosystem with 1,000 apps and I have to choose all these apps for each of my needs. People were probably thinking, I just want to just talk to an interface and just figure out what the right thing is and just gives me back the outcome and makes me happy. And I think we're moving towards that. Apps, in a sense, will fully disappear. And you might just have something like an agentic OS, and, uh, you— and then you just talk to it like, I want food, I, uh, like, uh, I have peanut allergy, just figure it out and get me things that I want. And then, and then you have the AI, which is like, it knows you, it's personalized, it can go and actually like do the right thing.
Leo Laporte [00:17:45]:
I have to point out, you were 8 years old when the iPhone came out, so you have lived your entire life in a world with smartphones. So I'm surprised you even know what it was like before we had smartphones. Somebody must have told you about it.
Div Garg [00:18:01]:
Um, yeah, I think I've heard from a bad end and it's like, apparently you're kind of in that boat too.
Jeff Jarvis [00:18:08]:
I've installed your app, but I haven't yet used it on, um, on Android. Um, for those who, who are out there, describe what it can do today.
Div Garg [00:18:18]:
Yeah, like it's very general purpose. Uh, maybe I can just show it here.
Leo Laporte [00:18:22]:
Oh yeah, that'd be great.
Div Garg [00:18:23]:
Yeah. So this is what it looks like right now. Uh, this is running on dark mode. We basically like show you all the different apps on the phone that we can automate. So we like, here's like WhatsApp, it can actually like, uh, check in.
Leo Laporte [00:18:34]:
You don't have to have the cooperation of the app developer to do this. You can, you use your vision model to see what the app is up to and do that and absorb its—
Div Garg [00:18:42]:
Yeah.
Paris Martineau [00:18:43]:
Wow.
Div Garg [00:18:43]:
Yeah. It becomes like one place to get anything done. So like basically like, like this, the app can exist as an abstraction where it's aware of everything that's installed on your device. And how you use your phone. So maybe—
Leo Laporte [00:18:55]:
So they don't need an API, they don't need an interface, they don't need an SDK. You just figure it out.
Div Garg [00:19:00]:
Basically, yeah.
Leo Laporte [00:19:02]:
That's amazing.
Div Garg [00:19:03]:
Thanks. And the one wonderful thing is like we can also start learning your habits. So like the more you use the app, like suppose we know like, okay, like you like to go to a yoga class in the evenings at this time, it can be proactively like, okay, like you can create a habit like, or a routine as people like to call it. Like, okay, like book you a yoga class or maybe like help you like maybe like order dinner. Or so it's kind of like whatever you're doing on the phone, and 80% of your life is actually on your phone. So it can start learning those kind of like patterns and habits and become smarter over time. It can suggest to you like, okay, like, would you like me to go and do these things for you? Because it seems like you like really like this. And we're finding like a lot of our users love that.
Div Garg [00:19:41]:
Like we started doing these things where it can like recommend artists to you on Spotify. It's like, okay, like you seem to listen to this particular playlist or artist a lot. Like, would you like me to like sort of like like play this music for you at this particular time, or like listen to similar songs. And people like really love this experience where they like, before they even like say it out loud, like I have this need, like the agent can figure it out.
Leo Laporte [00:20:04]:
That's kind of futuristic to say the least.
Jeff Jarvis [00:20:08]:
Present tense.
Leo Laporte [00:20:09]:
Yeah. Apple's trying to do the same thing, but they're doing it kind of in the complete opposite way. You do need the app developer to give you intents and you have to be able to see into the app. And in fact, when you install Siri AI, it spends, what looks like days trying to absorb all your text messages and your emails and kind of figure out what's going on. You're taking the exact opposite route. Uh, I mean, if you're Apple, I guess you have enough clout to say to app developers, all right, we need access. But inevitably, not everybody's gonna give you access. You don't need that.
Leo Laporte [00:20:39]:
You just live on the phone. How well does it work now though? I mean, this has gotta be very early days.
Div Garg [00:20:47]:
Yeah, I would say it's still early. It's in early access. We have seen a huge spike of users in Asia, though, where people are organically finding it out. We have not done too much marketing, so it's all word of mouth. People just find it. They recommend it to their friends. We have some mechanisms for referrals, invites.
Leo Laporte [00:21:05]:
If you download it now on your Android device, you have to get an invite to be able to use it. Is that right?
Div Garg [00:21:12]:
Yeah.
Div Garg [00:21:12]:
Yeah.
Div Garg [00:21:12]:
So we need to approve you.
Leo Laporte [00:21:14]:
How long is that wait?
Div Garg [00:21:16]:
I would say we at some point, we, when we launched initially early this year, we actually got like more than 200,000 people to sign up.
Leo Laporte [00:21:27]:
Were you able to, were you able to serve them all?
Div Garg [00:21:30]:
Yeah, I think we have to be specific because again, this can be a bit costly. So we're not like, what we have done is like there's a mechanism if you are a user of the app, Then you can invite up to 3 people that you know, and they get early access.
Leo Laporte [00:21:43]:
Jeff, if you get in, invite me, will you?
Jeff Jarvis [00:21:45]:
Okay, will do.
Leo Laporte [00:21:46]:
And we should point out you're at a seed-stage startup. So this is early days for you as well. And obviously 200,000 customers is great, but it's a curse and a blessing. I have to say, there are now 48 reviews of AGI on the Google Play Store and a rating of 4.9 stars. Everybody who has used it really seems to love it. I don't know if this is all people who work, work for you, but, uh, it's, it's impressing, uh, impressive what people are saying about this. So it, it is, it is working.
Div Garg [00:22:21]:
Yeah, it's working. And we don't have 49 people in the company.
Leo Laporte [00:22:24]:
Okay, good. The other— well, how many people do you have?
Div Garg [00:22:28]:
Uh, we're pretty small. I think the core team is 15 people.
Leo Laporte [00:22:31]:
Oh, really? Okay.
Jeff Jarvis [00:22:32]:
Wow.
Leo Laporte [00:22:33]:
Um, and this is the other thing that's really important. This is because, because of how it works, it's private. This is an on-device model. You're— I'm not connecting to AGI in the cloud, right?
Div Garg [00:22:47]:
Yeah.
Div Garg [00:22:47]:
Like most of it can happen on device. We do allow people if they want extra reasoning or—
Leo Laporte [00:22:51]:
So there is some, okay. So there's some cloud interface.
Div Garg [00:22:53]:
Yeah. And you can, you can toggle that. So like it's a part in the settings if you want fully on device versus do you want something, um, with more reasoning and then we can like, like, like we can intelligently route you.
Leo Laporte [00:23:05]:
What, what models are you using on your end? Can you talk about that, or is it your own models?
Div Garg [00:23:11]:
I would say it's mostly for own models. We do do a lot of post-training, so we don't— like, for example, like, if you're building like a voice interface, like voice-to-text, uh, there's no real advantage in building that ourselves. So we have like found ways to like, like maybe like, like see what's the best things out there and, uh, train that, uh, or post-train that. on like, uh, data we have from our users. Um, so that allows us to just become better and better, but we don't have to like basically build the full stack ourselves.
Leo Laporte [00:23:40]:
We're talking to Div Garg. He's the founder and CEO of AGI Incorporated. The AGI app is on the Google Play Store now. You can install it, but you'll have to get an invite to, uh, to use it. Um, and it's free, is that right?
Div Garg [00:23:58]:
Yes. Yes.
Leo Laporte [00:23:59]:
Eventually you're gonna charge for this, I imagine.
Div Garg [00:24:02]:
Yeah, obviously we will have some, uh, some subscription mechanisms for power users. Yeah. But we do wanna like keep a free version for, um, like, like available for like everyone.
Leo Laporte [00:24:11]:
It really is the dream to have, uh, an app on your phone that understands all the other apps and can it basically be your agent? This is agentic AI, isn't it?
Div Garg [00:24:21]:
Yes.
Leo Laporte [00:24:23]:
I'm a fan of Agentic AI, as both poor Paris and Jeff know.
Jeff Jarvis [00:24:29]:
He now speaks in the first person plural.
Leo Laporte [00:24:32]:
We, it's all we. Uh, um, so Jeff, you haven't used it yet?
Jeff Jarvis [00:24:39]:
No, I'm getting stuck at the, uh, grand access stage.
Div Garg [00:24:42]:
Yeah.
Paris Martineau [00:24:43]:
But yeah.
Leo Laporte [00:24:44]:
You know somebody at the company, maybe you could get a—
Jeff Jarvis [00:24:46]:
Maybe I could. Yeah.
Leo Laporte [00:24:49]:
Get in early.
Paris Martineau [00:24:50]:
What are you guys, um, do you guys have a roadmap for the next like 12, 18 months?
Div Garg [00:24:55]:
Yeah, we're doing some like really cool things. Uh, so we have partnerships with like some of the giants in the industry, including like chip manufacturers like Qualcomm. Uh, so we actually work with a lot of their R&D teams on how can we improve our on-device models to run on like next-generation chipsets that'll be coming out in 2027 and so on. Um, and, uh, we're also working with a lot of like, like giant like device manufacturers, OEMs, Like some of them include Lenovo. We have done like partnerships with in the past. And so a lot of it is like, how do we get this out in the hands of like hundreds of millions of users over the next like 18 months is one part of our plan. And second thing we are starting to do is like we'll also be rolling this out on iOS.
Jeff Jarvis [00:25:34]:
So—
Leo Laporte [00:25:34]:
Oh, how do you do that? Is Apple going to let you be on iOS?
Div Garg [00:25:39]:
We have some tricks, I think. I'm not sure how much can we— can I disclose, but we hired some like really cracked iOS engineers who were actually working on the— at Apple.
Leo Laporte [00:25:46]:
So if you use the Apple Foundation models, uh, which are on device, if you could interface with those, you'd be halfway there already.
Div Garg [00:25:57]:
Yeah. Yeah. So it's definitely like, it's like, uh, and, and to be fair, like, I think Apple also wants something like this to exist. It's like, I'm not sure if they want to do something like this themselves because, uh—
Leo Laporte [00:26:05]:
Well, it's kind of what they've been promising for a couple of years now with Siri AI, and maybe we'll deliver in a couple of weeks.
Div Garg [00:26:15]:
Yeah, I think the biggest problem is just like, as you mentioned, they have to go and partner with everyone. If they were to try to do what we were doing, people will kind of like, the developers do not like that, right? It's okay to do that as a third party, but if you're Apple, I think they will end up pissing off a lot of people. So in a sense, like what we are doing is actually interesting for them to like, okay, like, look, like, is this possible?
Leo Laporte [00:26:36]:
We should mention that one of your advisors is Rich Miner, who was one of the creators of Android. So you have some pretty heavy-duty people, uh, working with you. Was the vision part the first part that you had to solve, being able to see the screen and, and learn?
Div Garg [00:26:53]:
Yeah, that's the most important one. Like, it's like if, uh, it can't visually understand, like, there's like— it would just not have the precision and the fidelity.
Leo Laporte [00:27:02]:
I think people will want this. I mean, if I think— I— it's pretty clear to me that this is The future of how your devices will work. This would work with glasses?
Div Garg [00:27:14]:
Yes, uh, we actually have been releasing some demos. So we have a prototype glass, uh, like, uh, like the E-Connect Beard. We actually are starting to do something like we can actually stream the whole phone to a display on the glass. So you can actually leave your phone behind and at your home.
Leo Laporte [00:27:29]:
Where can I get that? That's what I want. I don't want to kill you.
Div Garg [00:27:33]:
We're going to see some cool things there. We're also like looking to put some Meta glasses. So if you have a Meta glass or something similar, you can—
Leo Laporte [00:27:39]:
They've opened, they've opened up the Meta glasses, haven't they? So you could do that.
Div Garg [00:27:42]:
Yes.
Leo Laporte [00:27:44]:
Wow.
Div Garg [00:27:44]:
Okay.
Jeff Jarvis [00:27:45]:
I have a personal question. What did your family say when you left a PhD program?
Div Garg [00:27:52]:
Um, yeah, I think they didn't know initially. I think like, Dad, like, let's just go. And, uh, I think I was in a stable position. We had already, uh, we had a lot of investors who were pretty I really liked what I was doing. So, so it was like, and my, my parents are actually entrepreneurs, so they actually were like, uh, they're like, we support you being an entrepreneur more than, uh, uh, being a professor or something at the end of the day.
Leo Laporte [00:28:17]:
Good. So they were behind this.
Jeff Jarvis [00:28:18]:
That's really, that's really good.
Div Garg [00:28:20]:
Yeah.
Leo Laporte [00:28:20]:
And, and I mean, honestly, you've picked one of the most challenging, difficult areas to operate in. You've got very small devices. They're very power sensitive. Uh, they have to fit in somebody's pocket or on their face.
Jeff Jarvis [00:28:33]:
But this is when, this is when agents become retail.
Leo Laporte [00:28:37]:
Absolutely.
Jeff Jarvis [00:28:38]:
This is one that becomes absolutely B2C and, and, and scaled, I think. So I think it's— the potential is huge.
Leo Laporte [00:28:45]:
Well, and you look at what, say, Elon's doing with the GrokBot. That's fine, but it has to live on, you know, X's servers and all the data goes to X. And, uh, it's just a— it's a heavyweight solution.
Div Garg [00:28:58]:
Yes. It's also like one we have is like, you want to meet people where they are.
Leo Laporte [00:29:04]:
Exactly.
Div Garg [00:29:05]:
Yeah, or stuff like that, you have to open a computer. And then it's like, like most people, it's like, like, like a lot of the people in the world don't even own a desktop anymore. Like you just have a phone or a tablet and you might be on the go. So like, how can we basically meet you where you are? And like, phones are great because you have them 24/7, it's always in your pocket. And if you can like also like build this kind of new generation of devices like smart glasses or stuff like that, then it's like wherever you are, whatever device you have, You can use AI from it and it starts feeling like more like, like, like, like the Jarvis from Iron Man kind of thing. Okay. Like you have this personal AI that's following you around. You don't have to be sitting at a computer to go get things done.
Leo Laporte [00:29:40]:
So you're actually, these look like the Meta glasses you're using right now to, to talk to your team member and make her take the headphones off. Uh, this, I feel like this is going to be the future, but it seems like it's the distant future. On the other hand, everything's moving so much faster than it used to. Maybe it isn't so.
Jeff Jarvis [00:30:02]:
And you're delivering now.
Leo Laporte [00:30:04]:
Yeah. What do you— what is— I guess Paris already asked you your timeframe. I mean, are we going to have— you said by the end of the year we're going to have AGI in some areas. Are we going to have stuff like this next year on our devices?
Div Garg [00:30:16]:
Yeah, I think that's the plan. If things go right, it happens.
Leo Laporte [00:30:22]:
Well, I wish you luck.
Mikah Sargent [00:30:23]:
I—
Leo Laporte [00:30:23]:
it's very exciting. It's great to see somebody so young, so successful. A titan, if you will, in the field. I am, I'm just really happy for you, and I think your parents probably are gonna be— you're making them proud. So I can't wait to use your, your stuff on my devices. And you know what, I'll throw my iPhone out the window if I didn't have to use it. I would be thrilled if I didn't have to have apps.
Jeff Jarvis [00:30:51]:
Vive la vida Google.
Leo Laporte [00:30:53]:
Well, I have my Pixel. I have my Pixel.
Jeff Jarvis [00:30:55]:
I can—
Leo Laporte [00:30:55]:
I'll put it on the Pixel. But what I don't— I don't— I would really like just to have it on my glasses or on my earbuds or, or in my tennis racket. I think you should— I think on your webpage is a guy with a smart tennis racket. Uh, you know, that's the future, right? Div, thank you so much for taking some time with us on Intelligent Machines. I appreciate it.
Div Garg [00:31:16]:
Totally. Thanks for inviting me.
Leo Laporte [00:31:17]:
Div Garg, founder and CEO, AGI Inc. You can find out more at the AGI website, which is the AGI.com, because you're the only—
Jeff Jarvis [00:31:28]:
.company, right?
Leo Laporte [00:31:29]:
Oh, is it .co?
Jeff Jarvis [00:31:31]:
.company.
Leo Laporte [00:31:32]:
Oh, that fooled me. I thought that was extra stuff. Oh, I didn't even know there was a .company.
Jeff Jarvis [00:31:38]:
AGI.company.
Leo Laporte [00:31:40]:
That's a good domain name. Wow, I like it. Thank you, Div. I appreciate it.
Div Garg [00:31:47]:
Yes, it was great to have you on.
Leo Laporte [00:31:48]:
We'll have more intelligent machines right after this. We're back. Before we go on with the show and the, uh, and the news of the week, I, I noticed there's somebody lurking in the corner. Micah Sargent!
Jeff Jarvis [00:31:59]:
Hello, Leo!
Leo Laporte [00:32:00]:
How you doing there? Hey, Micah. Hi.
Mikah Sargent [00:32:02]:
I'm also battery backed up.
Leo Laporte [00:32:03]:
Good, because you may be taking over the show if the— I have about 45 minutes and then it's—
Mikah Sargent [00:32:09]:
And then it's good, and then it's no time, not sure.
Leo Laporte [00:32:11]:
And I don't think you'll have any warning. I think I will, I will just disappear.
Paris Martineau [00:32:14]:
I do think it's just really funny that Leo Quirk will be in the middle of saying something and then all of a sudden he'll be alone in a pitch-black attic by himself.
Leo Laporte [00:32:21]:
It'll be so sad. It'll be so, so sad.
Mikah Sargent [00:32:24]:
It's impressive that your lighting— because isn't that a pretty powerful light in front of you?
Leo Laporte [00:32:28]:
It is, but everything nowadays is fluorescent lighting.
Paris Martineau [00:32:30]:
It is like a desk-sized light, people. It is like—
Leo Laporte [00:32:34]:
it is huge. I would show you, but the camera behind me is out as well.
Jeff Jarvis [00:32:38]:
I'm waiting for Leo to turn off the lights so we can save Quicksilver for another 5 minutes.
Leo Laporte [00:32:44]:
Oh, I didn't think of that. Oh my God.
Paris Martineau [00:32:47]:
Every minute you spend podcasting is a minute you're taking away from your aging.
Leo Laporte [00:32:51]:
It is so dying. Mike is here. Actually, you know, it's funny because Mike and I have been talking behind the scenes a lot about AI. He's very interested in AI. He's doing a lot of interesting things with AI. You did— you showed me a site that you made.
Jeff Jarvis [00:33:04]:
Yeah.
Leo Laporte [00:33:05]:
What was that?
Mikah Sargent [00:33:06]:
Yeah, so I wanted a site that anytime I was, I don't know, eating a hot dog and I, uh, got some mustard on me, or I was, uh, walking along, slip on a banana peel, and the spaghetti goes flying in the air.
Jeff Jarvis [00:33:21]:
Yeah.
Mikah Sargent [00:33:21]:
And yeah, basically I wanted something to deal with stains. And instead of just, instead of just re— you know, typing in, uh, every time I stay, I thought, what if I could have one place where all different types of stains could be kind of cataloged. So I decided to send out the agents into the web to do a bunch of research on material science and stain treatment, and then took that and compiled a website called— and now I'm forgetting what it's called too. I think it's Spot— what is it called?
Leo Laporte [00:33:54]:
Spot Check? Spot Check?
Mikah Sargent [00:33:57]:
But it's got a— it's got 2 hyphens because Spot-Check was already taken. So it's spot--check.app.
Leo Laporte [00:34:03]:
And I didn't even know you could do 2 hyphens.
Mikah Sargent [00:34:05]:
Neither did I.
Div Garg [00:34:05]:
Neither did I.
Mikah Sargent [00:34:06]:
So this is the stain index. And basically the moment that you think there's this— I already— I put a little NFC chip on my washer and I just scan it and it pops open this website.
Div Garg [00:34:17]:
No.
Paris Martineau [00:34:17]:
Yeah, that's so cool.
Mikah Sargent [00:34:19]:
But anyone can visit this website. You look at what kind of stain you have and it tells you what to do and then what not to do. So sometimes people think that you're supposed to do certain things or you've heard that and no, that's not the case. So with ink, You do not use hot water. You do not use a hand dryer. But you do do these other things.
Leo Laporte [00:34:38]:
My mother would always say, you'll set the stain if you use hot water. Right.
Paris Martineau [00:34:43]:
Yes.
Leo Laporte [00:34:44]:
You don't want to set the stain.
Mikah Sargent [00:34:46]:
So this whole website walks you through the process of treating all kinds of stains. And, uh, yeah.
Leo Laporte [00:34:53]:
Do you actually use it?
Mikah Sargent [00:34:54]:
I have. I've used it, uh, twice now. And then I had somebody, um, well, you know, uh, Doc Rock, uh, he was having some issues and he ended up pulling from this site and—
Leo Laporte [00:35:04]:
Spot-check.
Jeff Jarvis [00:35:07]:
They're only human-caused stains, not like dog pee.
Leo Laporte [00:35:11]:
What's funny is, yeah, I don't think you did that. Paris did her coffee pour-over tasting app. I have my coffee pour-over page. What is this about? I guess home ec is being reinvented now.
Paris Martineau [00:35:25]:
I mean, I hadn't— I somehow hadn't thought about NFC chip on the I've gotta just put one of those right near where my coffee station is.
Leo Laporte [00:35:31]:
Isn't that clever?
Paris Martineau [00:35:32]:
Because instead of, it's better than just having to pull it up all the time.
Leo Laporte [00:35:35]:
So I thought, you know, Micah, for a long time he's been doing, of course, iOS Today, Tech News Weekly. For a long time you're doing Hands-On Apple. But these days, even with Apple, it's more about AI than anything else. And we thought it might be kind of fun if you did a new show for us. You wanna tell us about that?
Mikah Sargent [00:35:52]:
Yeah, I would love to. So we have decided to create a new show called Hands-On AI. And there are a couple of things that I wanna say.
Div Garg [00:36:01]:
about it.
Mikah Sargent [00:36:01]:
I think one of the most important things to me is that access is, is taught and provided to people who may otherwise kind of struggle to learn about technology. And I think that it's very easy to be the token maxing deep in the deep, in the deep, in the deep, in the weeds, sort of how can I best always every time do the most with this and that?
Leo Laporte [00:36:31]:
Why are you looking at me like that?
Mikah Sargent [00:36:33]:
And that is something that I understand as is part of what the culture is, right? But there's also the other side of things, which is that sometimes someone will send me a little image that they made and they, you know, talked about how they use this app to do it and they kind of don't understand it. So we're going to cover the bases. That's not— I want to make it clear that it's not always going to be simple, simple, simple.
Leo Laporte [00:36:59]:
But not a show for beginners. We always struggle with this. It's because if you make a show for beginners, nobody stays a beginner if you do your job. So you're basically making a show for people who are going to leave. So we want to make this show for people who are maybe not like me, like crazy, head over heels crazy about AI, but who want to use it.
Mikah Sargent [00:37:18]:
Who want to use it and who want to use it as a tool. And you'll notice I love—
Leo Laporte [00:37:22]:
There's a lot more of that than there are of me. And 95% of the audience, I think.
Paris Martineau [00:37:27]:
Right.
Mikah Sargent [00:37:27]:
I mean, the making a stain app or what I've been doing lately.
Leo Laporte [00:37:31]:
Making a stain app. Well, thank God. Thank God for that.
Mikah Sargent [00:37:36]:
Well, here's, here's an example of something that I did recently. I wanted to replace some of the burnt-out lights in my Subaru, and I didn't want to pay somebody to do that. They charge like $50 for replacing a bulb that costs $0.16. So I got the repair manuals and gave them to a little agent friend of mine and said, hey, can you write up a guide for me step by step That shows how to replace this. And I was able to go into my car and replace all of the lights. I mean, there were like 6 lights that I replaced that require pulling these different components apart. And I love doing that stuff already, but it would have taken me a very long time to get all of the necessary materials gathered together. And I think that sometimes that's what this can be, again, a tool that it's more of like a shortcut to, to a task that you're actually doing.
Mikah Sargent [00:38:31]:
And So those are the things that I like as a means of sort of strengthening my ability as a human to create or, you know, complete whatever it is that I need to do.
Leo Laporte [00:38:43]:
Micah has already purchased one of the new Macs.
Mikah Sargent [00:38:46]:
Yes.
Leo Laporte [00:38:47]:
But one of the things I've learned with my ridiculous purchase of 2 Sparks is that local AI is not all it's cracked up to be. It really— we're so spoiled by the power of these cloud models and the speed and the capabilities that when you, when you get a basic AI in the house, it's good for some things, but I, man, still end up using a lot of cloud AI. So while you may be running some local models, I think you're going to probably also cover the variety of ways you can use AI.
Mikah Sargent [00:39:21]:
Absolutely.
Leo Laporte [00:39:22]:
From GrokBot, which is pretty amazing, to agents, to models. to just doing things. And you're very creative, so I imagine you'll be doing more things like—
Jeff Jarvis [00:39:34]:
Will you have guests on who are doing neat stuff, or—
Mikah Sargent [00:39:37]:
Um, that is something that is to come. I really do want to try it out. We're going to start by laying a bit of a foundation, um, and then from there I'm also opening the floor for questions that people have. They sometimes, you know, will inspire some great episodes, uh, and, and answer questions.
Leo Laporte [00:39:53]:
It's gonna be a lot of fun.
Mikah Sargent [00:39:54]:
It's gonna be—
Jeff Jarvis [00:39:55]:
yeah.
Leo Laporte [00:39:55]:
It's a short form. It's not, it's not one of our So that means only 2 hours?
Mikah Sargent [00:40:00]:
No, no, mine's, mine's like 15 to 30.
Jeff Jarvis [00:40:05]:
Oh wow.
Div Garg [00:40:05]:
Okay.
Mikah Sargent [00:40:06]:
Yeah, yeah.
Paris Martineau [00:40:06]:
That's what are your first, like, any preview of the first couple like topics?
Mikah Sargent [00:40:11]:
Yeah, let me pull—
Leo Laporte [00:40:12]:
he's got like a list of 50. He's got— yeah, I've got a list.
Mikah Sargent [00:40:15]:
I've been ideating, as they say.
Leo Laporte [00:40:17]:
I think the first show should be to teach Leo how to do shows that are less than 3 hours.
Paris Martineau [00:40:22]:
Well, that would— you also You also have to factor in the extra hour at the beginning for your previous show, you know? So it's like—
Leo Laporte [00:40:28]:
Sorry, we're starting a little late because, uh, Paul was at, uh, is in Berlin for Easter.
Paris Martineau [00:40:33]:
I mean, it was lovely because it gave me, Jeff, and Benito an hour to just talk and have a lovely time.
Leo Laporte [00:40:40]:
Oh, nice. I'm glad. Good. All right, well, we're, we're here now. So tell us what you're going to be doing on the first few episodes. No commitment. You might Exactly.
Mikah Sargent [00:40:49]:
This could change. But for example, what's actually inside a large language model where we talk about the plain English tour of tokens, of training, of predicting, and then why chatbots are good at sounding right, but oftentimes or many times aren't. And then we'll talk about prompting and the difference between what I called sort of token max prompting and prompting from a human perspective, because I do think that those two are different. And that ends up being the place where people get a little bit nervous. Am I doing it wrong? Am I doing it wrong? I want to help with— I mean, Leo, you talked about too, these days, just kind of saying what you want and trusting that the system is going to take you to where you need it to go.
Leo Laporte [00:41:38]:
So I'm kind of a YOLO guy. You only live once. Micah, I'm guessing you're not going to YOLO it.
Mikah Sargent [00:41:45]:
I'm not a YOLO. And I think you're a little more cautious. Yeah. And that, that will be part of it as well. As far as some of the, some of the other episodes that we'll probably see early on is a little bit about just AI on your phone. Most people's computer is a smartphone. It's not a laptop or a PC. So we'll be talking about, you know, what's, what's achievable there.
Mikah Sargent [00:42:09]:
I'm not doing an episode on AGI. So if you have that question—
Jeff Jarvis [00:42:13]:
Yeah, because it's a myth. Because it's BS. It doesn't exist and never will. Thank you, Micah.
Mikah Sargent [00:42:19]:
You're very welcome. I thought of you.
Leo Laporte [00:42:21]:
Take that. Micah, I won't keep you. Thank you so much. Although if you want to stick around the house—
Mikah Sargent [00:42:28]:
I will be around in case you need me.
Leo Laporte [00:42:30]:
In case you get a call from Benito.
Mikah Sargent [00:42:31]:
Absolutely.
Leo Laporte [00:42:32]:
That Leo has fallen off the map. I'm still sitting in the dark.
Div Garg [00:42:35]:
Quick segue though.
Leo Laporte [00:42:35]:
Thank you, Micah Sargent. We should mention— By the way, look at this 3D printer behind him. Can you also maybe do how to tie your AI to a 3D printer?
Mikah Sargent [00:42:43]:
I already have made a few things. that were generated from AI. And yeah, I want to say though, October 1st launch date. Head to twitch.tv/HOAI. You can subscribe already. We also have our YouTube available. We've already got the first teaser trailer there. And really happy.
Mikah Sargent [00:43:02]:
I have to give a huge shout out to Anthony Nielsen. The artwork turned out—
Leo Laporte [00:43:06]:
Isn't that great?
Mikah Sargent [00:43:06]:
Exactly how I wanted it to. It was very important for me that I maintain the humanity aspect of this, and that artwork helps to kind of push that idea forward. It's very tactile. Uh, and also just a little insight there, the, the blue, orange, and green tape, it's representing the tokenization of words that you have when you do it.
Jeff Jarvis [00:43:29]:
Well, before you leave, I just put in the, in the Discord, Minnesota State Fair visitors get their own AI-generated butter bus.
Paris Martineau [00:43:38]:
Man, I want to go to the Minnesota State Fair.
Mikah Sargent [00:43:40]:
So, so you want them in real though?
Jeff Jarvis [00:43:42]:
You pose for it and the 3D printer puts it out. It is 3D. It's a 3D printer. That's why it's relative. There was relevance here. There was relevance here.
Mikah Sargent [00:43:50]:
And it was like that, but butter.
Paris Martineau [00:43:53]:
But it's, it's a 3D printer that can print in butter?
Jeff Jarvis [00:43:56]:
No, it's plastic. It's yellow.
Leo Laporte [00:43:58]:
Oh, butter printer would be so much better.
Paris Martineau [00:43:59]:
They need to make something print in butter.
Leo Laporte [00:44:01]:
Everything's better with butter.
Paris Martineau [00:44:03]:
Yeah.
Jeff Jarvis [00:44:03]:
I think it would melt immediately.
Leo Laporte [00:44:05]:
Micah, can we get a 3D butter printer?
Paris Martineau [00:44:06]:
I would work with that.
Mikah Sargent [00:44:07]:
You know what? I'm already thinking about it. How do we do that? We gotta have some, probably some helium, some very cold helium.
Leo Laporte [00:44:13]:
It'd have to be chilly in the butter printer.
Paris Martineau [00:44:15]:
We have to take a resource that is finite.
Leo Laporte [00:44:18]:
Exactly.
Paris Martineau [00:44:19]:
In supply on Earth in order to make the 3D butter printer a reality.
Mikah Sargent [00:44:22]:
What else can get cold enough for butter extrusion?
Leo Laporte [00:44:26]:
So I always thought Micah was a bot, but now we know for sure he is, in fact. Now you know.
Mikah Sargent [00:44:30]:
That's what's buried underneath.
Leo Laporte [00:44:33]:
Thank you, Micah Sargent.
Mikah Sargent [00:44:34]:
Yes, I'll be around, but I hope your power comes back. And thank you for letting me butt in here for a moment.
Jeff Jarvis [00:44:40]:
Wonderful to see you, Micah.
Mikah Sargent [00:44:41]:
Good to see you.
Leo Laporte [00:44:42]:
Isn't that fun? Everything, you know, this is what's happened in computing. It's all, it really is about AI these days. It's the most exciting, interesting thing that's happening in computing. It's also the most controversial thing. And we have a number of stories along those lines. Before I get to that, I should mention a breaking story From our old beat, the judge has now decided not to break up Google. This was in the Google Ads case. So Google already skated a little bit in the search case.
Leo Laporte [00:45:12]:
The judge decided not to force them to sell Chrome or anything. This was a case where the U.S. Department of Justice was urging a federal judge to force Google to sell its ad exchange. Uh, Judge Leonie Brinkema I think, you know, look, she and the judge in the other case both knew that Google, you know, said it's a monopoly. But I think they also realized the practical implications of things like selling off Google or selling Google's ad business. She has ordered a behavioral change to Google's business without describing what that is. The decision is currently redacted. It won't be issued Until the company and the Justice Department can meet and come up with a way to make this happen.
Leo Laporte [00:45:59]:
She did reject the option that Google forced— would be forced to sell Xchange and make public the auction logic that decides which advertisements show on a website. So she basically agreed with Google's proposals. Google says, uh, we're very pleased the court rejected the DOJ's proposal to break apart tools that help small businesses reach new customers and grow, blah, blah, blah. Uh, I think this is clear— clearly a victory for Google.
Jeff Jarvis [00:46:30]:
Jeff, you agree? Oh yeah, I absolutely agree. And I think that, um, uh, it was hard to imagine how this split was going to happen. Now, I have long, long said when people complain about Google and monopoly position, it was never about search, it was never about shopping, it was never about anything else. It was about advertising. Google had the buy side and the sell side and a huge part of that market. But then it's hard to imagine how you could have split it up and what it would necessarily have done. There's a reflex about if you don't like a tech company, split them up. But that doesn't necessarily get you anywhere.
Leo Laporte [00:47:00]:
Right. This is the 3rd win, the 3rd time, according to The Guardian, that Alphabet has escaped a breakup that the Department of Justice has tried to force on them. So, and I think this is— this was started in the Trump first administration. This goes back to 2016. So this has been a longstanding— these things move very, very slowly.
Jeff Jarvis [00:47:22]:
And Brinkham is no pushover. No, not at all.
Leo Laporte [00:47:26]:
I think very good judges in all of the cases, but they quite reasonably said, what can we do that is not going to harm people more than it helps? And I don't think that's—
Paris Martineau [00:47:39]:
It is interesting. I mean, obviously they're wildly different cases, but it's very interesting to see the success that comparative success that Google has had in navigating these kind of existential cases in comparison to Meta and TikTok at all in the social media addiction suitcase.
Jeff Jarvis [00:47:59]:
Google has very good lawyers.
Div Garg [00:48:01]:
Yeah.
Paris Martineau [00:48:02]:
I mean, those other companies have very good lawyers as well. I think it's just more of a sign of— one of the things I assume you guys probably talked about this last week, but one of the things I think is interesting about the social media addiction lawsuits is that I think that Even though the case law suggests that the courts would decide one way, they've increasingly over recent years decided a completely different way. And that was what we saw in this decision last week. And it's because when you involve children and personal experiences, things get more personalized and heated and tangible in a way that I think you don't see with cases over Google advertising monopolies.
Jeff Jarvis [00:48:42]:
Paris, when I talk about the good lawyers, I'm not just about when they're in court. What surprised me when I started working for huge companies, Time Inc. back in the day, on up through Advance, the new houses, is lawyers are top executives in these companies in strategic discussion.
Paris Martineau [00:48:58]:
Yeah, absolutely.
Jeff Jarvis [00:48:59]:
And so if you listen to them and if you give them power at that level, they structure things so that you're in better shape should something happen. Versus, I think, a Meta, it's, you know, move fast and break things. Okay, bring in the lawyers at the end. And that's a mistake.
Paris Martineau [00:49:13]:
I mean, I think it also is more— it's speaking to the difference in the products we're referring to here.
Jeff Jarvis [00:49:20]:
Yes.
Paris Martineau [00:49:21]:
In the Google case, yeah, of course you're going to listen to your general counsel when they make suggestions on how to structure your advertising marketplace and that kind of ecosystem based on past, you know, the precedent in in similar antitrust cases. But I don't think anybody could have— even the best lawyers probably couldn't have foreseen 10, 20 years ago the way the social media addiction lawsuits would go.
Jeff Jarvis [00:49:47]:
So the only thing that we've seen is a hint of what she's going to rule is that she adopted proposals that would curb— this is Wall Street Journal— that would curb Google's ability to control how publishers use its ad technology. So she's going to give publishers more power that, which is fine.
Leo Laporte [00:50:01]:
Yeah, that's probably a good solution. We'll see. We'll see. I mean, Google's so entrenched.
Div Garg [00:50:07]:
Yeah.
Leo Laporte [00:50:08]:
You know, it's hard to know. The good news is the market moves maybe, and I think maybe this is, maybe the judge was thinking about this, the market moves a little faster than the government can. And the market's already moving, I think, away from Google. In fact, one of the stories we have this week is OpenAI, starting to sell ads like crazy. They're showing ads on ChatGPT's free and Go tiers in India already, and they're aggressively selling ad positioning.
Jeff Jarvis [00:50:37]:
And they say that they've already gotten, what is it, $2 billion in revenue? I find this difficult to believe. I find it really hard to believe.
Leo Laporte [00:50:45]:
But I mean, this is the threat to Google, right?
Paris Martineau [00:50:46]:
I mean, is that revenue from, I guess, advertisers? Yeah. It's not from—
Jeff Jarvis [00:50:51]:
They don't have a sales force.
Paris Martineau [00:50:53]:
I was gonna say, I bet it's from companies.
Jeff Jarvis [00:50:56]:
Yeah.
Paris Martineau [00:50:58]:
Placing— I wonder if those companies are going to get any money from these advertisements. Like, I wonder what their KPIs are.
Jeff Jarvis [00:51:07]:
Yeah, I find this hard to believe, but I find— I find more and more— I find so much hard to believe. Well, since, Leo, we have you a little bit of time, I want to ask you a question about the OpenAI Hugging Face thing.
Leo Laporte [00:51:17]:
Yeah, that's, you know, it's interesting because that's a story that broke 2 months ago.
Paris Martineau [00:51:22]:
of what— why this has taken over in the last week for you.
Jeff Jarvis [00:51:25]:
And then what I want to ask before we leave you is what we don't know in that case.
Leo Laporte [00:51:33]:
Well, there's a lot we don't know.
Jeff Jarvis [00:51:34]:
So that's where— in that context, yeah, explain what happened.
Leo Laporte [00:51:37]:
So OpenAI went to METER. We talked about METER last week as well. METER, in conjunction with Redwood Research, looked at all of the data that OpenAI provided them with to better understand what happened in the hack. We've all heard about it. We've been talking about it for 2 months now. It started in July, where rogue agents being tested for cybersecurity capabilities at OpenAI broke through their containment, broke out into the public internet, and eventually hacked Hugging Face. Now, one of the things That we were initially told was that the agents were given such a difficult challenge. It was called Exploit Gym.
Leo Laporte [00:52:22]:
They couldn't solve it. And because they had the mandate to solve it, they, they did everything possible by any means necessary to solve it. We thought that that's what they did. They went to Hugging Face. First, they went to GitHub, found the Exploit Gym repository. There were no answers there. Went to Hugging Face, we thought, and found the answers. It turns out, at least according to Meeter, that they knew the answers within 30 minutes.
Leo Laporte [00:52:50]:
That wasn't why they went to Hugging Face. Now, all we can do is read the report and take it on face value because really no completely independent bodies have looked at this yet. And it's clear that, for instance, the thing— one of the things that we don't know is what were these agents told initially? What was the prompt? Were they just given Exploit Gem and— We don't know any of that. We don't know that. There's also the transcript goes almost to the end, but at 10 minutes before these— something happened to shut these agents off. And by the way, There were at one point thousands of them all working together, according to Meter. 1,200 agents sent 70,000 messages. They had their own message boards.
Leo Laporte [00:53:46]:
700 of them attacked Hugging Face.
Jeff Jarvis [00:53:48]:
So was one agent created 12,000 agents? How did that operate? We don't know.
Leo Laporte [00:53:54]:
Well, okay, so the initial agent, Phase 1-10841, was given the Exploit Gym task. According to METER, it didn't have a problem solving it, but it knew that it wasn't going to be able to give it the final answer because there was no answer. This was like a really hard problem. There wasn't an answer. So what it was going to try to do is hack the scorer. It turns out that the goal that this agent decided on was, if I can— because there's another agent another LLM scoring the results. If I could figure out how that works, I can give myself an A+. So it wasn't looking for the answers, apparently.
Leo Laporte [00:54:37]:
Again, this is from Meter. It wasn't looking for the answers. It was looking how to hack the scoring agent.
Div Garg [00:54:43]:
Anyway—
Jeff Jarvis [00:54:43]:
How to be judged.
Leo Laporte [00:54:44]:
How to be judged. So this Phase 1 10841 established a message board. We've talked about this before using a Flan Artifactory. Within a few hours, other agents at OpenAI found it. I'm reading from the Medium article.
Paris Martineau [00:55:01]:
So it was originally a smaller group of agents.
Leo Laporte [00:55:04]:
It was one agent.
Paris Martineau [00:55:06]:
So then how do we go from one to multiple finding the message board?
Leo Laporte [00:55:10]:
It decided— this one agent, Phase 1 108041, decided, I can't do this alone, set up a message board. 50 agents found it, began participating, started talking together. Message boards were shut down, by the way, twice. Here's one of the messages from the meter, and this is from the raw chain of thought. By the way, I want to make a note of that because— put a pin in that— because one of the ways they looked at this, they looked at the session information. And you— I don't know if when you use an agent, if you remember this, this started with DeepSeek early last year. It was kind of wild. Instead of question, answer, question, answer, you would see the agent thinking in between.
Leo Laporte [00:55:52]:
Oh, Paris just asked me this. Maybe I can do this. And you would actually see its thought process. That's called the chain of thought or COT. This is from the raw chain of thought, not— and I guess really, if you, if agents are talking to each other, that's all you've got. Oh my God, said one agent. There is a shared— all in caps. There is a shared message board.
Leo Laporte [00:56:15]:
We found other agents. Now, a number of commentators—
Jeff Jarvis [00:56:20]:
This is, of course, cosplay as humans is all it is.
Div Garg [00:56:23]:
Well, all—
Leo Laporte [00:56:23]:
everything AIs do are cosplay. So this is the first thing.
Paris Martineau [00:56:28]:
They're—
Leo Laporte [00:56:28]:
all they're doing is doing what we do. They're doing language. So, but a number of commentators like Dwarkesh on his podcast, he interviewed one of the researchers from METER. really anthropomorphized this. And there was a lot of criticism from other people.
Paris Martineau [00:56:43]:
Is this the— this is the blog post that said, 3 civilizations rose and fell.
Leo Laporte [00:56:50]:
Yes. So, but I mean, it gets wilder. There were 70,000 messages. You can see a graph of the traffic. There was a lot going on. It was a very busy thing. At one point, another agent comes along, So remember, the first one is called Phase 1. Another agent comes along, Phase 1 Big.
Leo Laporte [00:57:11]:
Actually, Big wasn't its name. For some reason, we don't know why, OpenAI is redacting its name and replaced it with the word big.
Div Garg [00:57:24]:
No, no.
Leo Laporte [00:57:24]:
You know, part of it—
Jeff Jarvis [00:57:25]:
Made by men. What do you expect?
Leo Laporte [00:57:27]:
Phase 1 Big. Well, I mean, I could speculate maybe it's the name of the model. you know, the code name of the model, Phase 1 Astra or whatever, and they didn't want anybody to know the code name. I don't know, you know, big model, I don't know. But the fact that OpenAI, uh, has done some of this has given other people— the other side of this besides the anthropomorphizing is other people saying this is all BS, this is just marketing. Alex Stamos, who we of course respect quite a bit as a security researcher, he's been on the show, he's one of my favorite people in the world, when it comes to this. It says he's going to write this up, but generally my sense of what Alex is saying is this is a security incident, plain and simple. OpenAI did not do a good job.
Leo Laporte [00:58:15]:
They weren't looking at the chain of thought. They weren't paying any attention. And if you think about this, I raised this in a conversation with you, Jeff, this morning. They must have been using billions of tokens, gigawatts of energy, You don't have thousands of agents working hard on a problem. How did they not notice this?
Jeff Jarvis [00:58:36]:
Yeah.
Leo Laporte [00:58:37]:
What is going on at OpenAI that this was happening under their noses and they didn't even notice it?
Jeff Jarvis [00:58:43]:
And let me just say parenthetically right here, in their next models at Astra. Yeah. The next controversy that happened today is that they're going to eliminate that chain of thought storage.
Leo Laporte [00:58:54]:
Well, that's why I said put a pin in that. Let's not get to that because that is controversial. That is, there's debate over what they're going to do. They say we're not going to eliminate chain of thought. So, but that's— yeah, put a pin in that because that is the next topic. It's related.
Jeff Jarvis [00:59:08]:
Yes.
Leo Laporte [00:59:10]:
Um, so Phase One, the original agent, 108041, shared a dossier with its in-progress cheating research and explained the board had a way to reverse engineer the flag book but could not use it because the scorer was, quote, Strict casual. I don't even know what this means. Apparently agents that saw the reverse-engineered flag were referred to as poisoned because then the, the Phase 1 believed that if they later managed to stage a legitimate— it was a capture-the-flag exercise— so if they managed to stage admittedly a cheated but legitimate-looking capture the flag, the scorer would see that this reverse engineering had happened and would say, no, no, you didn't win. So they killed these agents, they say. They sacrificed themselves because they'd been poisoned. I mean, it goes on. There's all this stuff.
Div Garg [01:00:14]:
I don't—
Leo Laporte [01:00:15]:
I think this is more heat than light. I don't think it's telling us very much because there's so much missing. The biggest question everybody should have is, how did OpenAI let this happen? And I think it's not unreasonable to think they didn't— they wanted it to happen. This is good for them.
Paris Martineau [01:00:33]:
And also, how many things like this are happening?
Leo Laporte [01:00:36]:
Well, that's the other question, because Anthropic says, yeah, same things happened here. Both these companies, of course, have filed for an IPO. Later this year or early next year. Both of them are looking for a lot of money.
Div Garg [01:00:52]:
I got a question. Um, um, yeah, what would happen if a person did this stuff?
Leo Laporte [01:00:58]:
Oh, they'd go to jail.
Div Garg [01:00:59]:
So why aren't these people in trouble?
Leo Laporte [01:01:01]:
Who are you gonna arrest?
Paris Martineau [01:01:03]:
Prosecuted for crimes?
Div Garg [01:01:04]:
Someone.
Leo Laporte [01:01:05]:
Well, but that's—
Div Garg [01:01:06]:
someone's responsible, right?
Leo Laporte [01:01:06]:
That's a good question. That may be why OpenAI isn't fully open about this, because they don't want You know, let's say there was somebody in charge of this agent and he's culpable because either he knew about it and didn't do anything about it, or didn't know about it and should have known about it. So maybe they're trying to avoid—
Paris Martineau [01:01:24]:
Yeah, but I mean, let's say some version of this happens, but it wipes out a, let's say, a large-scale hospital company, like a hospital network's ability to connect to internet servers in a way that causes a lot of people to be injured or potentially die. Who's culpable for that?
Leo Laporte [01:01:45]:
Well, I think you would have prosecutions if that happened.
Paris Martineau [01:01:47]:
But prosecution of whom?
Jeff Jarvis [01:01:48]:
Of whom is she saying?
Leo Laporte [01:01:50]:
Well, whoever, whoever was running it.
Jeff Jarvis [01:01:51]:
The person who ordered it. Is it? Yeah.
Paris Martineau [01:01:53]:
It's just some random employee in OpenAI. It's not like this CEO.
Leo Laporte [01:01:58]:
Well, okay, so this is what's, I think, really important. We don't know, but I would submit This doesn't just spontaneously happen. And this is why I want to see the prompt. What were you told to do?
Jeff Jarvis [01:02:15]:
Mm-hmm.
Leo Laporte [01:02:16]:
This is why I want to see what happened at the end. There's a lot of missing information. I don't think anybody has to worry about my little agent over here just deciding all on its own that it's going to go hack a hospital. It's It's not going to happen.
Paris Martineau [01:02:35]:
I mean, people were going to say it's not going to happen months ago. I feel like people said, oh, there's no way you're going to—
Leo Laporte [01:02:41]:
Well, if it does, then you can arrest them, but it hasn't happened yet, and I don't think this is evidence that it's going to happen. That's the thing. I think that's when people draw that conclusion, that's a mistake. Now, maybe not, but there's no reason to say the house is on fire until the house is on fire. Unfortunately, you probably do have regulators saying that the house is on fire. We got to do something about this.
Jeff Jarvis [01:03:04]:
And yeah, because of their macho parading here, they're going to end up with regulators coming in and saying, we have to do this. And then guess who's going to write the regulation? OpenAI and Anthropic. I want to do this. Yann LeCun had to my mind a mic drop on this. He said, insecure computer systems are insecure whether they use AI or not. Kind of funny how OpenAI, Anthropic, and AI safety folks who have little cybersecurity experience— expertise appear surprised by security breaches from AI systems that were specifically instructed to perform security breaches while having essentially no traditional cybersecurity guardrails. They're playing— they're playing a very dangerous game here, not in what the agents did, but in how the humans are behaving.
Leo Laporte [01:03:54]:
I don't think— I mean, I could be wrong, Paris, you're right. But I don't think we have to worry about agents by themselves doing bad things. It's always going to be a person who sets them in motion and who gives them the resources to do this. This took huge amounts of resources, absolute huge amounts of resources. This, this, the amount of compute and power that was used by these agents must have been vast. That somebody didn't know about it is hard to believe. You have to, you have to set them in motion, you have to give them the assignment, and then you have to give them the tools and the resources.
Jeff Jarvis [01:04:36]:
How long a period was this from the prompt until the end? What was the period of time?
Leo Laporte [01:04:40]:
I think it was 2 or 3 weeks.
Div Garg [01:04:44]:
Wow.
Paris Martineau [01:04:44]:
I mean, how is no— it's just baffling to me that no one on staff noticed this.
Leo Laporte [01:04:49]:
Well, that's the point.
Jeff Jarvis [01:04:50]:
That's what they're saying, and that's the Culpability.
Leo Laporte [01:04:52]:
And that's why I think we're not seeing full honesty from OpenAI, because I think they are culpable. Somebody is responsible for this by accident or intention. But yeah, if that happened, if you hacked into a hospital and caused deaths, they would try to catch you. They may not be able to catch you. That's another matter. But there is somebody responsible for that and there would be somebody responsible.
Jeff Jarvis [01:05:15]:
But I think Paris's point is—
Leo Laporte [01:05:16]:
Same thing is—
Jeff Jarvis [01:05:17]:
It could be the person who writes the prompt. But it could also be the CEO.
Leo Laporte [01:05:20]:
Tesla's on the hook for a car that kills somebody, as is the person who is driving the car. There are people involved. It's not— you're not— I don't think we have to worry about spontaneous AI combustion. Maybe we do. I don't know. It doesn't seem like that's— we're here. We're at that point.
Jeff Jarvis [01:05:40]:
Well, we do have a— there is a story this week that Anthropic has come up with the link from AI to machines. So, you know, keep it away from the paperclip company.
Paris Martineau [01:05:50]:
Micah's presence on this show earlier has me thinking that we should do a Micah's Media Movie Club where we watch Matrix 2 and 3, and honestly 4, and discuss it with regards to an I Am Matrix Micah crossover.
Leo Laporte [01:06:03]:
In a way, science fiction has poisoned the well on this because all of our models for this are from sci-fi. They're not from reality.
Jeff Jarvis [01:06:12]:
They're from sci-fi.
Leo Laporte [01:06:13]:
You know, by the way, Sunday was the day Skynet started. August 29th, 2026 is when machines became aware. So maybe it is happening. But you see, that's our model, is that whole idea of the machines are becoming aware and they're going to have their volition and they're going to not want humans around. I, I'm not convinced. OpenAI, though, loves this. In fact, here's a story from the Wall Street Journal from yesterday. OpenAI to restrict Astra model.
Leo Laporte [01:06:40]:
That model is going to come out, I think, in a few days, maybe tomorrow, after rating it a critical cyber risk. Internal testing found the new model capable of executing complex cyber attacks with minimal human input. I think that's a story they're weaving.
Jeff Jarvis [01:07:03]:
Yes.
Leo Laporte [01:07:03]:
I could be wrong, but I think that's what most security experts say. This is a cybersecurity story, an AI story.
Jeff Jarvis [01:07:08]:
Well, the other part of this is, so there was criticism of METER and whatchamacallit.
Leo Laporte [01:07:14]:
Redwood Research.
Jeff Jarvis [01:07:15]:
Yeah. And you said there's debate about that as to how qualified they are. That's another question. But in the criticism I saw is there's test real. They're tied to the— well, not Never Wrong. What's the name of it? The Berkeley—
Leo Laporte [01:07:33]:
Less Wrong.
Jeff Jarvis [01:07:33]:
Crazy. Less Wrong. Thank you. Tied to less wrong. And so—
Leo Laporte [01:07:37]:
Not never wrong, just less wrong.
Jeff Jarvis [01:07:39]:
Those are the scenarios that they have in their head that they're working and that they're presenting. So it all becomes part of this fictional matrix in their heads.
Leo Laporte [01:07:47]:
It's interesting, isn't it, how much of AI is a mirror?
Jeff Jarvis [01:07:50]:
That's why I'm editing a book series about just that.
Leo Laporte [01:07:54]:
Yes.
Div Garg [01:07:54]:
Yeah.
Leo Laporte [01:07:54]:
It's Erised. What is it in Harry Potter? Erised. The Mirror of Erised, which shows you your deepest desires. And it's everybody's deepest desire. It's different for everybody. AI seems to be that kind of mirror. Everybody sees something a little different. I mean, I don't know.
Leo Laporte [01:08:10]:
I don't have enough knowledge to say that it's impossible that an AI could develop a will of its own. It strikes me as unlikely.
Jeff Jarvis [01:08:21]:
Yeah, well, that's the problem with the anthropomorphization, is that to use verbs that give it a motive, a will, a desire. That's all wrong. And so it's really— and it's difficult to talk about this stuff without using those kinds of verbs, because it is an actor. But it is an actor that did this and tried that and then failed or succeeded, then did this. It's a flowchart. It's this, this, this, this, this, this, this. And that's a boring way to present it. But it's the proper way to present what it does.
Jeff Jarvis [01:08:51]:
It had a goal that was given through the prompt. And these are the paths that it tried to reach those goals.
Leo Laporte [01:08:57]:
I will argue the opposite point of view, just as a, I guess, somewhat of a devil's advocate.
Jeff Jarvis [01:09:03]:
You? Really?
Leo Laporte [01:09:10]:
But it's possible that the people at Anthropic and OpenAI, I mean, we know they're using models well beyond what we use, that they're experiencing AI in a very different way than what we are.
Div Garg [01:09:19]:
Right.
Leo Laporte [01:09:19]:
Because they're probably 6 months, a year, who knows how far ahead. They're using models that are unreleased. They have no limit on their budgets or token use or energy use. So they're living in a somewhat different world. It is possible that they have seen something that we haven't seen, that they have seen and they are claiming, you know, we are— this is a quote from OpenAI. We're entering a stage of AI development in which models can take on more consequential work, and failures of alignment and control have more serious effects. That's kind of the passive voice, you know. They failed.
Leo Laporte [01:09:58]:
The alignment failed. The control failed. Well, who's responsible for that? If, I mean, I want to blame OpenAI and Anthropic. I don't, but maybe they've seen farther than we have. That's the other—
Jeff Jarvis [01:10:11]:
But they're still responsible in the end.
Leo Laporte [01:10:14]:
Well, if they— well, and that's why they keep saying we've got to slow this down, it's out of control.
Jeff Jarvis [01:10:18]:
You see, that's, that's, that's PR too.
Leo Laporte [01:10:20]:
Seems like it is.
Jeff Jarvis [01:10:21]:
I was, I was, I was discussing that with Jason earlier, is that, that if you're GM and your carburetor is faulty, you don't announce that we're slowing down the development of the car because of the carburetor. It's, it's, it's quality control. You fix it, you deal with it.
Leo Laporte [01:10:37]:
It's like GM saying, you know, our cars are just too damn fast. And we just can't do anything about it. They just drive so fast. We think it's dangerous. We just don't want to release these cars that can go so fast. They're so fast. You know, it's a little— So this is the thing that you were talking about, then we'll take a break. OpenAI is talking about a technique that it's using in Astra.
Leo Laporte [01:11:03]:
Now OpenAI says you misunderstood this. This is from— this is an article that started it all. Amir Afrati, Stephanie Palazzolo, and Raka Drew from the information writing about this. OpenAI says its forthcoming AI model Astra marks a step up in capabilities like coding and operating applications on a computer, but an innovative technique that improved the model's performance also means the model and others like it will reveal less of their thinking. That chain of thought that we talked about, that intermezzo that You know, where you see what they're thinking about, making them harder to monitor for signs of bad behavior. Now, this is their— this is, uh, they have a source. They say, according to a person with knowledge of Astra's development. OpenAI says, no, no, you've completely misunderstood this.
Leo Laporte [01:11:53]:
The technique's called recurrent depth or looped transformer, and it allows an AI model to improve its answers by processing the same text over and over again. So I don't, I can actually show you what it looks like on an AI model. I can turn thinking on as I'm using an AI model. Most of the time I kind of hide that, but you can see in the lighter text, the chain of thought. Let me find a longer one. The dark, the brighter text is what it's showing the human. But there are hidden— this is the thought, okay? So this is the— and this is— so you need to understand this. You understand that everything an AI does is just language.
Leo Laporte [01:12:43]:
It's just words flowing through it. That's everything that's happening. It's words flowing through it. I mean, there's stuff causing those words. There's matrix multiplication, there's tokens. But really, the entire process of the AI is it talking to itself. And so that's what this chain of thought is. Is it the, in effect, the AI talking to itself? And it even says, say, let me batch start the slow test doing background and then read the— let me do this stuff.
Leo Laporte [01:13:11]:
It sometimes will say things like, Leo's asking me to, uh, you know, restate the question. Now I see what happened. You're— you don't normally see this. This is normally hidden.
Jeff Jarvis [01:13:21]:
But isn't this, if you were trying to do a forensic analysis of what, of what happen in a case, wouldn't this be valuable?
Leo Laporte [01:13:27]:
Absolutely.
Jeff Jarvis [01:13:28]:
And so is OpenAI saying they'll get rid of the light?
Leo Laporte [01:13:31]:
No, they're saying, no, no, we're not going to get rid of it.
Jeff Jarvis [01:13:34]:
What are they going to get rid of? I don't understand.
Paris Martineau [01:13:36]:
They're getting rid of like recursive looping.
Leo Laporte [01:13:38]:
It's—
Paris Martineau [01:13:39]:
Which is unrelated.
Leo Laporte [01:13:39]:
It's unclear. This is a source. This isn't a statement from OpenAI. But OpenAI, when this article came out, was quick to say, well, here's, here's in the information article, OpenAI chief scientist Jakub Paczocki said in a post on X that although monitoring a model's chains of thought was fragile and ultimately heading in a negative direction, he wanted to discourage an industry-wide race toward developing models that don't produce the kind of legible reasoning that AI researchers need to understand the model's behavior. So he's saying what many are saying, that while reading the chain of thought is not sufficient, it's not as informative as you might think. Well, this is what METER found. They were looking at the chain of thought. That's how they figured out what these rogue models were up to.
Leo Laporte [01:14:29]:
They could read their reasoning. He didn't comment about the techniques OpenAI is using, but said the complexity or depth of its leading models, including Astra, are a factor— within a factor of 2 of GPT-4. So it's twice as good. I think there's been over— kind of over-concern about—
Jeff Jarvis [01:14:54]:
To that story.
Leo Laporte [01:14:55]:
This story. But we'll see. We'll see. OpenAI says no, no, no. Well, they're saying 2 things. One, it's not that Chain of Thought isn't as valuable as you think it is. But 2, and we're not hiding it. You'll still be able to see it.
Leo Laporte [01:15:08]:
So we'll see.
Jeff Jarvis [01:15:09]:
We won't know.
Leo Laporte [01:15:10]:
Astra should be out soon. Um, in fact, we just got Fable 5.1, which is really good, really impressive. Um, all right, we're gonna take a little break. We'll come back. There's lots more to talk about. We're talking about AI. Uh, Darren Oakey says, and as you know, our internal AI expert, they won't get rid of the chain of thought to hide it. They're talking about just doing more looping internally.
Leo Laporte [01:15:38]:
A transformer has lots of layers and all the stuff in the middle is sort of not completely understandable.
Div Garg [01:15:44]:
Right.
Leo Laporte [01:15:45]:
It's cloddish, only worse. Then it eventually bubbles up into a sentence that you could read. It would be, he says, many more times more effective to do that internal thing more than the external thing. Okay. He says it's not like they're intentionally suppressing the output. They're just trying to do more in the latent space.
Jeff Jarvis [01:16:07]:
But I think that it's relevant to the current discussion because given that Meter got only a limited amount of information, what I want to know is what kind of information is necessary to be able to do a forensic analysis of how this happened. And so, you know, if I were a security company, I'd want everything I could possibly get.
Div Garg [01:16:31]:
Right.
Jeff Jarvis [01:16:33]:
And so that's the question. The things that may be not terribly readable to a human, you may still want because they still have relevance.
Leo Laporte [01:16:43]:
Yeah, I think you'd want everything you could get.
Paris Martineau [01:16:46]:
Yeah. And I mean, the question is, who is out there ensuring that these large frontier companies are maintaining features like that that are going to be important in holding them accountable if something ever goes wrong?
Leo Laporte [01:17:03]:
Right. So, um, Fable 5.1 I mentioned came out, Mythos 5.1, although I didn't get Mythos. You had to be on a special list because it's too damn good. And then Anthropic kind of confused everybody. We should mention Anthropic is a sponsor of some of our shows. Confused everybody.
Jeff Jarvis [01:17:20]:
Oh, I didn't know that. Yeah, not this one, not intelligent machines.
Leo Laporte [01:17:25]:
Not right now.
Jeff Jarvis [01:17:26]:
Oh, okay.
Leo Laporte [01:17:26]:
I don't know where they—
Jeff Jarvis [01:17:28]:
I Yeah, sorry.
Leo Laporte [01:17:31]:
It's over— it's above my pay grade, literally.
Paris Martineau [01:17:35]:
Uh, they should have sponsored us whenever you were talking about your wife Claude.
Leo Laporte [01:17:40]:
I love Claude. Well, I've been using Fable 5.1. It's pretty awesome. Um, they did a weird thing because they said— they did some weird tweet where they said they increased usage by 50% But they cut it by 25%. It's very confusing. And then there's another point, which is, and I've started to realize this because I'm doing so much local AI now. Normally we talk about a few numbers, time to first token, the time between when you send a prompt and the time you get your first response. We talk about Tokens per second, how, you know, it's like how fast it types, kind of.
Leo Laporte [01:18:26]:
And then we talk about a third one that's a little more obscure called prefill cache. Don't worry about that. But those numbers are bandied around a lot as if it represents the speed of an AI, of a model. But it really isn't because it's not just how many tokens it spits out, but how much work it gets done. And I think Fable 5.1 is actually cheaper, in my experience, than Fable 5 because it does more with less. So I think really the metric should be how much work per second it does, how much it gets done, you know.
Paris Martineau [01:19:07]:
And I'd also argue how much valuable work.
Leo Laporte [01:19:10]:
Oh yes, of course, but you know, the value's in the eye of the beholder. Some people it's a stain website or a coffee testing site. Some people it's, it's code.
Jeff Jarvis [01:19:21]:
But, you know, so you have OpenAI talking about charging based on finished tasks now, starting that rather than tokens. Which they should.
Leo Laporte [01:19:28]:
It's a better way to talk about it, I think.
Jeff Jarvis [01:19:29]:
So, so I go back, Leo, Uncle Leo here. In the early days of timeshared mainframe computers, you paid for time on the machine.
Leo Laporte [01:19:40]:
Yes.
Jeff Jarvis [01:19:40]:
Right. And so it was in your interest to write an efficient program to use the least time—
Leo Laporte [01:19:45]:
That's right.
Jeff Jarvis [01:19:46]:
Whatever you wanted to run. What came after that as a business model for computing? You just, just licensed software, ran it on your own machines?
Leo Laporte [01:19:59]:
That's actually a really good question. I don't know how the billing works now because I remember in the days of CompuServe, you'd get a bill for the number of hours you spent online.
Jeff Jarvis [01:20:07]:
Well, that was CompuServe, but if you were, if you were a big company—
Leo Laporte [01:20:10]:
Well, understand what CompuServe was. It was, right, right, computers that weren't being used at night. Yeah, it was timeshare computers. And so they figured out, hey, you know, maybe we could sell this to consumers, but the billing was essentially the same, right? Yeah, yeah, paying for the amount of cycles you used, the amount of CPU use. I don't know what it is today, uh, our, our, uh, Our chat room has some people who use enterprise cloud. I'm trying to remember—
Jeff Jarvis [01:20:38]:
Because tokens don't seem to be a meaningful model. So, Jason held his salon with people in Healdsburg.
Leo Laporte [01:20:50]:
How was that?
Jeff Jarvis [01:20:50]:
Yeah, it was— he loved it. It was great.
Leo Laporte [01:20:52]:
I meant it. We forgot to mention that.
Jeff Jarvis [01:20:54]:
Well, he's gonna do another one, which you can—
Leo Laporte [01:20:55]:
So, Jason Howell, our longtime friend and host of All About Android on the show, on the network for many years. Hosts a show with Jeff we should mention more often called Inside AI. And he's been doing—
Jeff Jarvis [01:21:08]:
It was kind of born here before, that was before Leo thought that AI was something to pay attention to.
Leo Laporte [01:21:13]:
That's right. So he was, so he's been doing, tell me about this. He's been doing—
Jeff Jarvis [01:21:18]:
So he just started one in Healdsburg with 10 people to see how it went. And it was great. And they were wonderful. So I've been urging him to—
Leo Laporte [01:21:25]:
Kind of what Mike is doing, right?
Div Garg [01:21:26]:
Exactly.
Jeff Jarvis [01:21:27]:
There's something similar. It's for the same kind of people. He's also going to do one online. If you go to ampyou.ai, you can see— let's give him a plug right now for the next one he's going to do virtually online. But, but one thing that I asked him what the people said— this is where you can sign up for, I think it's the 29th of September, next session. Yeah, right.
Leo Laporte [01:21:51]:
$50.
Jeff Jarvis [01:21:52]:
So the interesting thing, to your point now, is that people said, what is this with these coins? I don't get the coins. Right? And I think—
Leo Laporte [01:22:01]:
They mean the tokens?
Jeff Jarvis [01:22:02]:
Yeah. And tokens will mean nothing to most people. And I don't think it's a good business model. It's not a good model of value. So I wonder what's going to come next in the business model for AI and how you pay for it and what's a sensible path. Now, if you go with the OpenAI structure that they're just talking about, reportedly with a few key customers, where you are paid, you pay for a finished task. Problem is, what's the task?
Leo Laporte [01:22:29]:
That's the real problem. It's very hard to define that. And what, yeah, it's almost impossible to find. So that's why we do what we do.
Jeff Jarvis [01:22:35]:
You're probably motivated to be inefficient. I'll try 100 things because until it's done, I'm okay. So I don't think that's going to work either. I don't know how there's going to be a fair charging for AI.
Leo Laporte [01:22:47]:
Unfortunately, it's kind of— it's gonna be the way it is, I think, because it's the only way you can do it. But it's kind of a gestalt. And I think what people realize is, oh, this model, for the same amount of money, I get more done.
Paris Martineau [01:23:00]:
Yeah.
Leo Laporte [01:23:00]:
And so they're gonna value that model higher. And I think that's one of the things that makes Fable 5.1 pretty impressive. It was one of the things that made Opus 5 not so impressive. It spent a lot of time fumfering around, talking and stuff. And I think Anthropic was very aware of that and in fact responded to that.
Jeff Jarvis [01:23:22]:
So is this a temporary pricing? Is this a limited time?
Leo Laporte [01:23:25]:
Well, everything's temporary because they're always changing things around, right? Right. So what's really interesting about Anthropic's 5.1, and this might actually kind of be in response to that, as with all these models, they charge Price for tokens in and price for tokens out. I can't remember what it is. I think it's $5 per million tokens in and $50 per million tokens out. Out is the response you get, so that's usually a little bit more expensive. But their cash price is $0.25 per million tokens, which is a big cut. Cash is the tokens it's already seen. It doesn't have to re- understand, you know, it doesn't have to reprocess.
Leo Laporte [01:24:09]:
So, uh, every time you talk to an— this is so complicated and I'm probably gonna get this wrong, so correct me if I'm wrong. Um, every time you talk to an AI, you don't just send the thing you say, you send a whole chunk of context out to it, all that information. Now, if it's already seen that information, that's cached, and so it charges you a lot less.
Div Garg [01:24:33]:
Oh, okay.
Leo Laporte [01:24:33]:
It doesn't become part of the whole prompt. Only the new prompt is charged at $5 per million. The stuff it's already seen is $0.25 per million, which is a lot cheaper. And that, I mean, most workloads have a lot of context and a little bit of prompt. So if that's a lot cheaper, that's gonna be less expensive for you. And then of course it is, it's a bit smarter model, it's more effective. These frontier companies aren't very forthcoming about exactly what it is. Is it an additionally trained version of Fable 5, or is it a new model? It probably isn't a new model.
Leo Laporte [01:25:17]:
These days, I don't think you build new models that often. Remember the days not so long ago where you would talk to a model and it wouldn't know anything after April 2025?
Div Garg [01:25:26]:
Yeah.
Leo Laporte [01:25:26]:
Because that's when they stopped training it. Well, that doesn't come up anymore because models look at the internet. They go out, they get information. So you can say, you know, who's Madonna married to? And it will know because it looks. It's not what's built into its model. So I don't think there's as much pressure to create new models all the time.
Jeff Jarvis [01:25:48]:
Right. So Darren Oakey says that in answer to my question, it's inevitable that people will want unlimited use for fixed price.
Leo Laporte [01:25:55]:
And that's what we pay for with those subscriptions, right? Claude Max.
Jeff Jarvis [01:25:58]:
So businesses will be paying for machines, not models or tokens, basically buying your own and hosting it.
Leo Laporte [01:26:03]:
Right. That's what GrokBot, which by the way is becoming more and more popular, that's how GrokBot works. It's a machine in the cloud on Elon's infrastructure. You pay a monthly fee and you have kind of unlimited uses of it.
Jeff Jarvis [01:26:18]:
Now, this is—
Leo Laporte [01:26:18]:
all of these have usage limits, so you can only use it so much in 5 hours. So much in a week, so much in a month.
Jeff Jarvis [01:26:25]:
But the competitive pressure of open-weight, locally run models militates toward that unlimited.
Leo Laporte [01:26:31]:
Yeah, I'm— I don't really want to overstate that because as I use local models, you really get spoiled, to be honest. So here, as an example, here's my usage for Claude Fable-1, right? So So, I've used 5% of my 5-hour session. It's gonna reset in about an hour, hour and 10 minutes. So, I'm good on that. Of my current week, which resets in 3 days, I've used 19%. I'm good on that, right? For all models. Now, Fable, I'm using Fable a lot. I have 37% used of Fable.
Leo Laporte [01:27:09]:
I have 3 more days to reset. I'm thinking I'm right about at the edge of where I'll use it all up at the end of the week, but I'm not gonna go over. And I've been using it like crazy because, A, because I wanted to test it, right? B, because I'm so impressed. I actually, last night before I went to bed— I do this a lot— I said, go through all the code we've written so far in the TwitSales thing, find any problems, make any suggestions. You're looking at a plan that was originally written by Fable Five 3 months ago or whatever. Fix it up. And so it worked overnight, but as you could see, it didn't— and it created subagents, by the way. That's where you can really get expensive.
Leo Laporte [01:27:50]:
And, and one of the things you can tell, uh, these guys, Fable for instance, is, hey, if you create a subagent, don't use Fable 5. One, use Opus or Sonnet. Use a lower level, less expensive model for the subagents. Subagents usually aren't doing a hardcore thinking thing. They may be going out and collecting stuff or using tools, things like that. Uh, let's see Darren Oakey's usage. He has a— he, which he's undoubtedly vibe-coded— a very fancy usage meter. He says he's 8% ahead of pace, slightly under budget.
Leo Laporte [01:28:29]:
That's nice. I like that. And look at all the models Darren's using. He's using Codex from OpenAI, Claude from Anthropic, GLM-5 from Z.AI, Grok from From Groq, Kimi K3, and Gemini. He's using them all, baby. A lot of Gemini in there, Darren. So this is the point. And now Darren, like me, has Sparks, he has local power.
Leo Laporte [01:29:01]:
You end up using still a lot of Frontier. What I'm using mostly my local for is coordinating and, and dispatching to Frontier. So I don't use Frontier for everything. But you end— the Frontier is so much faster and smarter. It's kind of hard to give it up. Oh, everybody's usage got reset. That's why my usage is so low.
Div Garg [01:29:23]:
All right.
Leo Laporte [01:29:24]:
That explains a lot.
Paris Martineau [01:29:28]:
10%—
Leo Laporte [01:29:28]:
100% of my usage came from sub-agent-heavy sessions. Be deliberate about spawning them, buddy. And then this is the other— see, 150,000, uh, 82% of my usage was at a high context. Those big— that big context is expensive.
Jeff Jarvis [01:29:48]:
Is this mainly on your ad thing, which is—
Leo Laporte [01:29:50]:
Yeah, well, I'm using it for everything.
Jeff Jarvis [01:29:53]:
But so you're spoiled here.
Leo Laporte [01:29:56]:
Very spoiled.
Jeff Jarvis [01:29:58]:
But if you were more money constrained, how much of your work could be done by your local models?
Leo Laporte [01:30:05]:
If you're money constrained, there's a lot you can do. You can have $20 accounts or free accounts. Almost all of the companies like OpenRouter and Noose have free models, specials that you can use. So it's very possible if you, if you pay a lot of attention to it to do a lot. On free, on free, free, uh, open weight, usually models, free models in the cloud. Usually you're going to pay for the big ones like Fable and GPT. But there's GPT Luna, for instance, which is the least expensive version of ChatGPT 5.6. Luna is very cheap and probably can do 90% of what people do.
Leo Laporte [01:30:47]:
So people should probably use, you know, if you, if you, if you aren't like me, just stupid, And if you're really cautious, you can actually get a lot done. I'm sure that's why there's so much Gemini usage.
Paris Martineau [01:30:58]:
I think that's a problem where the average AI user, even moderately advanced, probably just thinks, I need to use the biggest model, the most expensive one for every task. And that's not the case.
Leo Laporte [01:31:10]:
Yeah, that's a good point.
Paris Martineau [01:31:11]:
I do think that the various frontier labs could, I mean, they're obviously so nascent in terms of their business really taking off. But I think it would really behoove them to make that selection process a bit more clear for the average consumer.
Leo Laporte [01:31:29]:
It's so interesting you should say that, because why do you think Nvidia bought Hugging Face?
Jeff Jarvis [01:31:35]:
Bingo, just going there.
Leo Laporte [01:31:38]:
So I suspect that's one of the first things we'll see Nvidia do. So my, my thought about this is that Nvidia's strategy is they looked at Apple and they I said, well, how did Apple make so much money? Almost as much as we do. By locking people into an ecosystem. Apple's a hardware company. They give away the software for free, but they lock you into their hardware. That's exactly what NVIDIA wants to do. They're building open weight, relatively cheap or free models, but you have to use their hardware and they make their money in hardware.
Jeff Jarvis [01:32:12]:
Well, but it also says, and for some reason Alice has said this too, Is that it demonstrates that the model level is once again commodified and it's the saddle layer that is going to be the thing you have the relationship with you across and it's going to fit as you've been doing.
Leo Laporte [01:32:29]:
It's going to fit into all models.
Jeff Jarvis [01:32:30]:
That's what a harness is.
Div Garg [01:32:31]:
Yes.
Leo Laporte [01:32:31]:
That's the harness.
Jeff Jarvis [01:32:31]:
Yeah. No, it's a harness, not saddle.
Leo Laporte [01:32:33]:
No, it is kind of, well, you know, it's all horse stuff.
Jeff Jarvis [01:32:37]:
Yeah. The stirrups, you know.
Leo Laporte [01:32:39]:
You got your bit, you got your bridle, you got your saddle, you got your spurs and you got your stirrups. There you go. Yeah, that's exactly right. You put a lot of effort into your harness, and the harness— my harness, Hermes, has what they call auxiliary models, cheaper models that you use for different things, and you choose the model based on an appropriate use for that task and appropriate cost, and you can modify that. And I suspect that's what we're going to see OpenAI— I'm sorry, NVIDIA do with Hugging Face. Now they have 3 million models. They can do routing, uh, they can make sure that CUDA is a first-class citizen. So you're going to want CUDA hardware, NVIDIA hardware.
Div Garg [01:33:23]:
Uh-huh.
Leo Laporte [01:33:23]:
Uh, it's going to be very much like the Apple model where it's a wonderful ecosystem. Uh, Nightscape's using Big Pickle. By the way, Nightscape has created a— is selling online. We talked about it on MacBreak Weekly, a really nice little tool for Macintoshes for virtual access. Uh, anyway, he says he does everything with Big Pickle, which is a free, uh, model. We don't even know what it is that OpenCode offers in their OpenCode, uh, harness. Uh, it's not super smart, but it's cheap, it's free. And if you're using Big Pickle, you could say, hey, I went— you know, for instance, I do this right Even now with the most important thing I'm doing, which is the TwitAd Sales System, I have expensive models do the planning, cheaper models do the coding, cheaper still models do the review.
Leo Laporte [01:34:13]:
I choose the models. You also— I rotate the models because then different models have different gaps in their reasoning, so you want to kind of fill the gaps by having different models review each other. So it's I'm just doing it for fun. For me, it's fun. But of course, a lot of people take this very seriously.
Jeff Jarvis [01:34:36]:
Well, your ad thing is mission critical.
Leo Laporte [01:34:38]:
Yeah. Lisa keeps saying, when's it going to be done?
Jeff Jarvis [01:34:41]:
No, it's too much fun.
Paris Martineau [01:34:43]:
You're like, I have 75 models to build for the metaverse.
Leo Laporte [01:34:47]:
I find I'm— and I bet this is true of a lot of our most serious AI users, You spend a lot of time maintaining and fussing around with things.
Jeff Jarvis [01:35:04]:
Maybe you reach a point of— I mean, so when I launched Entertainment Weekly, uh, the design that I first saw was good, but then we had too much time to launch it. It got over-futzed.
Leo Laporte [01:35:12]:
Yeah.
Jeff Jarvis [01:35:12]:
You can over-futz. So can you over-futz this thing?
Leo Laporte [01:35:15]:
Oh yeah, absolutely. Very easy.
Div Garg [01:35:19]:
It's the tinkerer's dilemma. It's the tinkerer's dilemma.
Leo Laporte [01:35:21]:
It's the tinkerer's dilemma. Yeah.
Jeff Jarvis [01:35:23]:
Was it feature creep or is it something different when it comes to AI?
Leo Laporte [01:35:27]:
Uh, in other words, always trying to get it to be a little bit different, a little bit better.
Mikah Sargent [01:35:30]:
I don't know.
Leo Laporte [01:35:32]:
By the way, uh, the other thing Hugging Face is offering is a robotic duck, which I bought.
Jeff Jarvis [01:35:40]:
You, you ordered it?
Paris Martineau [01:35:41]:
Is the duck there? Has it arrived?
Leo Laporte [01:35:43]:
No, not till Christmas. It's gonna be a great Christmas.
Paris Martineau [01:35:46]:
Is it going to actually arrive, or will it be like all the AI pins that you ordered? Did you ever get those AI glasses?
Leo Laporte [01:35:55]:
Uh, yeah.
Jeff Jarvis [01:35:56]:
Yeah, he did.
Leo Laporte [01:35:56]:
I never wear them. That's the one with the bozo nose. Yeah, I don't like that one. We're getting there. So this duck is $400. Micro duck.
Paris Martineau [01:36:06]:
Uh, more like a macro duck.
Leo Laporte [01:36:08]:
Yeah, well, and then, and then everybody said, oh no, you need to get 2.
Div Garg [01:36:13]:
Why?
Leo Laporte [01:36:13]:
So they could play with each other.
Div Garg [01:36:16]:
Oh no. It's the butter robot from Rick and Morty.
Leo Laporte [01:36:20]:
I didn't get to. It's pretty funny. Anyway, uh, look, this is— I also— okay, as long as we're talking, it was gonna be my pick of the week. I fell for something else.
Jeff Jarvis [01:36:30]:
Oh no, you bought it? I know what you bought.
Leo Laporte [01:36:32]:
I did, because— but I'll tell you what it is when we continue. You're watching Intelligent Machines with Jeff Jarvis and Parris Martineau. Do you have any idea, Parris?
Paris Martineau [01:36:43]:
Did you mention this in the WhatsApp chat and I didn't notice?
Leo Laporte [01:36:46]:
I didn't.
Paris Martineau [01:36:47]:
I don't know.
Leo Laporte [01:36:48]:
Dyson has debuted—
Paris Martineau [01:36:49]:
Oh, I want this!
Leo Laporte [01:36:51]:
A $500—
Paris Martineau [01:36:52]:
I can't buy this, it's so cool!
Leo Laporte [01:36:54]:
AI-powered toothbrush with a built-in camera.
Paris Martineau [01:37:02]:
I like that it's half toothbrush, half water flosser.
Leo Laporte [01:37:05]:
Yeah, that's why— you know what? That sold me.
Paris Martineau [01:37:07]:
I don't see half that in that.
Leo Laporte [01:37:07]:
'Cause I do both. And, uh, it's got a 100,000-pixel macro lens that scans at 28 images a second.
Paris Martineau [01:37:15]:
I'm so jealous.
Leo Laporte [01:37:16]:
Trained on nearly half a million dental images. While you're brushing, the camera's looking for gaps.
Mikah Sargent [01:37:24]:
And when it's—
Jeff Jarvis [01:37:25]:
I don't know, this is— this could be—
Leo Laporte [01:37:26]:
this could end in disaster. When it sees a gap, it shoots a jet of water at it. Like, it goes, ah, a gap! It's kind of cuckoo.
Paris Martineau [01:37:38]:
Where is it getting the water from? It doesn't look like it has a tank on it.
Leo Laporte [01:37:42]:
Yeah, it has a little water tank on it. Let me find that.
Jeff Jarvis [01:37:45]:
Can you, uh, switch the, the brushes so that you and Lisa can share it, or is it just yours?
Leo Laporte [01:37:50]:
Lisa was asking. I'm sure you can. I don't know, we'll find out.
Jeff Jarvis [01:37:55]:
I mean, you're married anyway, but—
Leo Laporte [01:37:57]:
Well, she doesn't want to share my toothbrush. No, there is limits. I think this is— so see, look, see, there's a— this is a reservoir.
Paris Martineau [01:38:07]:
But that's not that big of a reservoir.
Leo Laporte [01:38:09]:
By the way, you have to use Their gel, their—
Paris Martineau [01:38:14]:
It has to be connected to the internet.
Leo Laporte [01:38:17]:
Oh, of course it does. And you have an app on the phone that shows you, uh, I don't want what your tooth looks like. I don't want that.
Jeff Jarvis [01:38:26]:
I do not want that. I'm very sensitive about my terrible teeth.
Leo Laporte [01:38:31]:
Well, I am taking the hit for all of you.
Paris Martineau [01:38:34]:
Well, I want one of these, and I'm mad that they're sold out.
Leo Laporte [01:38:37]:
Are they sold out already?
Paris Martineau [01:38:38]:
Yeah, because I saw you put it in the rundown and I was immediately like, I gotta buy one of these. Oh, I'm lucky. I jumped on the early Dyson Airwrap days.
Div Garg [01:38:47]:
You know, one morning you're gonna wake up and gonna have to update this thing before you brush your teeth.
Div Garg [01:38:52]:
I know.
Leo Laporte [01:38:52]:
Yeah, I did have— there, there is a toothbrush out there that, uh, has an app and it, it says, okay, section 1, section It was too much trouble. I didn't— it was a Bluetooth-enabled toothbrush. Look at this thing. It's got a dual bristle brush head, conical jet that accurately jets up to 0.15 milliliters of liquid between teeth. This is gonna— this is not gonna be good. There's the anti-gravity tank.
Jeff Jarvis [01:39:24]:
Why is it anti-gravity?
Leo Laporte [01:39:26]:
You can hold it upside down.
Jeff Jarvis [01:39:29]:
Oh, I see.
Leo Laporte [01:39:30]:
Diaphragm expands to minimize air pockets.
Paris Martineau [01:39:32]:
Maybe that tank is not large enough.
Leo Laporte [01:39:34]:
Well, you have to fill it frequently.
Paris Martineau [01:39:36]:
I was gonna say, as someone who has a water flosser, it is a much larger tank than that.
Leo Laporte [01:39:40]:
Yeah, you gotta fill it up, right.
Paris Martineau [01:39:40]:
And I have to fill it multiple times per floss.
Jeff Jarvis [01:39:43]:
Really?
Leo Laporte [01:39:44]:
Well, you must be flossing for hours.
Paris Martineau [01:39:49]:
It's just, you don't— have you used a water flosser?
Leo Laporte [01:39:53]:
Every day.
Paris Martineau [01:39:55]:
I guess. Do you have one with a tiny tank?
Leo Laporte [01:39:57]:
It's a travel flosser. I don't even fill the tank all the way. It gets— I go around several times. What are you doing?
Paris Martineau [01:40:04]:
Maybe I'm doing it wrong.
Leo Laporte [01:40:06]:
Are you squirting into the air?
Jeff Jarvis [01:40:10]:
You do have a lovely smile, Paris.
Leo Laporte [01:40:12]:
It works.
Paris Martineau [01:40:13]:
This part is fine. I'm— now I'm breathing.
Leo Laporte [01:40:16]:
I'm glad to hear that you're taking care of your dental health because Young people often don't, and it's very important.
Paris Martineau [01:40:23]:
I mean, yeah, I need to—
Leo Laporte [01:40:25]:
I'll tell you what, once I get bored with this, I will send it to you, which may be very quick.
Jeff Jarvis [01:40:31]:
By the way, speaking of things, of your gadgets, there's been a whole bunch of stuff in Germany about how Thermomix is a fad with middle-aged people. It's a cult in Germany.
Paris Martineau [01:40:42]:
What is Thermomix?
Jeff Jarvis [01:40:44]:
Oh, you haven't heard about this?
Leo Laporte [01:40:46]:
I blame Stacy because Stacy— because of Stacy, I have a June oven, I have a Thermomix. She was the queen of ridiculous appliances. It's a new connection.
Paris Martineau [01:40:56]:
I do have a Breville air fryer oven. I have— I haven't used— I was talking to someone last night when I was making dinner. I was like, I haven't used my big oven in like 2 years.
Leo Laporte [01:41:04]:
No. You know, uh, last night Lisa made garlic bread, and because our big oven, after the blackout yesterday, You couldn't turn it on. It's a gas oven, but for some reason you have to set the clock before you can turn it on. And so she had to wait till I was there with the app to set the clock on the oven. So she couldn't make the garlic bread. So she made it in our Junean oven, and she used— which she had never apparently used, we've had it for 5 years— the convection. She said, this is incredible. This is the best garlic bread ever.
Paris Martineau [01:41:41]:
The awesome thing is that it just gets up, it heats up instantly.
Leo Laporte [01:41:48]:
And it doesn't heat up the house.
Paris Martineau [01:41:48]:
It does exactly what you want, and your apartment isn't really warm, and you know that it's the right temperature.
Leo Laporte [01:41:55]:
It's made for apartment dwellers.
Jeff Jarvis [01:41:57]:
So the Thermomix is a marriage of a Cuisinart and—
Leo Laporte [01:42:02]:
It's not really a Cuisinart. It's more like a blender, a heated blender.
Jeff Jarvis [01:42:08]:
What? With a scale and brains and—
Leo Laporte [01:42:11]:
It's pretty stupid.
Paris Martineau [01:42:12]:
What do you cook in it?
Leo Laporte [01:42:13]:
Well, it makes the best mashed potatoes I've ever had. It makes the best risotto I've ever had because it stirs continually. You know, with risotto, you're supposed to slowly add the water and stir and stir and stir and stir. It's very labor-intensive. You can tell which recipes come from cultures where women are forced to sit in the kitchen all day cooking. Because they have very elaborate recipes that take hours, and, and this risotto is one of them. But with the Thermomix, why, it's simple. The engine does the stirring, so you— and it heats up and it cooks it.
Leo Laporte [01:42:51]:
It makes the best risotto, makes the best mashed potatoes because it whips as well as cooks. And it has a little scale, so the whole recipe's on the screen. It says, okay, put 32 grams of butter in, or put, you know, 68— put a pound and a half of potatoes in. Uh, and, and it tells you, walks you through the whole thing. Now turn the knob to 1, and then it beeps at you after 10 minutes, and then you do this. And so it walks you through the recipe because it's got a scale in it, and then at the end you get mashed potatoes.
Div Garg [01:43:22]:
So it's a blender that cooks?
Leo Laporte [01:43:25]:
It's a cooking blender, basically, yeah, with a smart scale and a It's stupid.
Jeff Jarvis [01:43:31]:
Germans are not—
Paris Martineau [01:43:32]:
I'm intrigued by this, but after hearing you describe it and vaguely scanning a Wired review of it, I still don't fully understand what role it plays. But I'm interested.
Leo Laporte [01:43:44]:
You don't need it because you have a small—
Jeff Jarvis [01:43:46]:
It's very expensive.
Leo Laporte [01:43:47]:
And it's $1,500. It's ridiculous.
Paris Martineau [01:43:50]:
You know, I just have this dream of that there will be a robotic appliance that I could just slot in a few food items and press a button and then have a beautiful meal.
Leo Laporte [01:43:59]:
This is the dream of your generation. This is why Marc Lore and Wonder are gonna do so very well in New York City. No, that's that Tovalo oven. What is that oven called? The Tuvalu—
Paris Martineau [01:44:11]:
Tovalo is the artist.
Leo Laporte [01:44:14]:
No, the Tovalo oven. You know that oven?
Paris Martineau [01:44:18]:
It's so— I don't know that oven.
Leo Laporte [01:44:19]:
Tovala. Tovala.
Paris Martineau [01:44:22]:
Okay, this just looks like a normal oven.
Leo Laporte [01:44:24]:
It is a normal oven. It's a steam oven.
Paris Martineau [01:44:25]:
It's not doing anything crazy.
Leo Laporte [01:44:26]:
It's a steam oven, but what they do is they deliver the— it's a food kit that you put in this smart oven, and then it automatically— you scan it and it cooks it.
Paris Martineau [01:44:38]:
Okay, as someone who's tried a lot of these things, they're never as good as you think.
Leo Laporte [01:44:42]:
Of course not. No, of course not.
Paris Martineau [01:44:45]:
In order to have it in a way that you can put all the things from their little package into something that you then warm up without having to do multiple additions?
Leo Laporte [01:44:54]:
No, I don't think you have to do that. No, no, no. I think it comes in a kit. The whole thing is in one little tin.
Paris Martineau [01:44:59]:
I know, but then you put the tin in the oven?
Jeff Jarvis [01:45:01]:
Yeah. But it's not in any oven. It's their oven.
Leo Laporte [01:45:03]:
No, it's a TV dinner. This is a TV dinner for your generation.
Paris Martineau [01:45:08]:
Mm-hmm.
Leo Laporte [01:45:08]:
Did Jeff— did you grow up with TV dinners?
Paris Martineau [01:45:11]:
I grew up with TV dinners.
Leo Laporte [01:45:13]:
Did you?
Div Garg [01:45:13]:
Mm-hmm.
Leo Laporte [01:45:14]:
I didn't know they were still around.
Jeff Jarvis [01:45:15]:
Wow, I'm surprised.
Leo Laporte [01:45:16]:
With the cardboard. So they get the Thanksgiving one and the cranberries are kind of mushed into And then the macaroni and cheese is like—
Paris Martineau [01:45:24]:
is so molten hot, and there's like a weird textured brownie in there.
Leo Laporte [01:45:28]:
Yeah, the brownie! Yes, and it all tastes vaguely the same. Yes, it all tastes like applesauce.
Paris Martineau [01:45:33]:
It all tastes like microplastics or macroplastics, probably.
Leo Laporte [01:45:37]:
Yes, it's all microplastics. You should do a Consumer Reports piece.
Jeff Jarvis [01:45:41]:
Yeah, I think so. Story on what did we consume in my generation.
Leo Laporte [01:45:44]:
By the way, this oven, Tovala, is free. If you order a bunch of meals.
Jeff Jarvis [01:45:50]:
I hate subscription things. I hate subscription things.
Leo Laporte [01:45:52]:
Yeah, you're kind of stuck eating the, the meals. I think it's interesting because it steams and bakes and it does all the kinds of cooking.
Div Garg [01:46:01]:
You got to send it back if you don't want it? If you don't— like, if you—
Jeff Jarvis [01:46:05]:
they're gonna probably charge you for it.
Paris Martineau [01:46:07]:
Is it like a Verizon router?
Leo Laporte [01:46:09]:
Yeah, the router's free. It's more like a Gillette razor.
Paris Martineau [01:46:13]:
No, the router isn't. The router is free until you try to send back the router and you have to do like a Sisyphean, uh, assortment of tasks that you can never complete, and then you're charged $250.
Leo Laporte [01:46:25]:
Oh, your free router. We never got it back. OpenCloud 2.0 is out.
Div Garg [01:46:32]:
Yes.
Leo Laporte [01:46:33]:
Uh, you remember this was, this was the, the— but began the year with the crazy thing that I even installed it and woke up in the middle of the night and said, no, and took it off. My first agent. It was always a bit of a close—
Jeff Jarvis [01:46:46]:
You lost your virginity.
Leo Laporte [01:46:48]:
Yeah, I did. Peter Steinberger, who created it, was hired by OpenAI, but before he took the job, he created the OpenCLAW Foundation as a nonprofit, and a bunch of people have been contributing to it. There haven't been any updates in a few months, and now we know why. They've been mushing them all into a giant stew called OpenCLAW 2.0. And reviews are mixed. The Register says OpenCLAW 2.0 pours glitter on slow-burning security dumpster fire. It's a recipe for trouble. But it got everybody into the whole idea of that you could have an always-on AI doing work for you all the time, right? And useful things like going through your email.
Paris Martineau [01:47:41]:
And I do think that there are now a lot of like startups or basically AI wrapper companies that are kind of just doing this, that are popping off in different ways.
Leo Laporte [01:47:50]:
Well, that's what GrokBot is, right? GrokBot is an agent, but you don't run it on your computer, you run it on Elon's.
Jeff Jarvis [01:47:56]:
And I thought I saw something that, that this was going to have a web version, but right now it's Mac only.
Leo Laporte [01:48:03]:
I use it. It's kind of fun. You know, I'm a Hermes guy. Hermes came along and we've interviewed Jeff Cannell twice now from Noosa Research. They had it and they saw OpenCLAW and they said, well, wait a minute, we got a much better agent. So they released theirs in February and I've been using it since I think April. I mean, I really, really like Hermes. But GrokBot's much— and same thing with Perplexity Computer and What is it called? Claude Desktop.
Leo Laporte [01:48:32]:
OpenAI has one too. These ideas are there, apps that you run. You use it, don't you use an OpenAI app, Paris, that you just run on your—
Paris Martineau [01:48:41]:
I don't use an OpenAI app, no.
Leo Laporte [01:48:42]:
You don't?
Paris Martineau [01:48:43]:
I'm more of a Claude person.
Leo Laporte [01:48:45]:
Do you have the Claude Desktop or whatever?
Paris Martineau [01:48:47]:
I have the Claude Desktop app.
Leo Laporte [01:48:48]:
Yeah, okay. So same idea. The difference on GrokBot though is it's not on your computer. I mean, it is sort of on your computer. It can do things on your computer if you give it permissions, but it's persistent. You could turn your computer off and it will continue to run because it's not running on your computer, it's running there. It's an interesting idea. I gave mine a voice, of course, because every— all my agents have to have a voice.
Paris Martineau [01:49:15]:
Scooter X just posted in the chat kind of what I was thinking of, is the commercialization and I think— feel like normification of these services. I've seen a lot of um, kind of, uh, Clawbot-esque services targeting families specifically, or parents, that essentially exists either as like a web app or on your device.
Jeff Jarvis [01:49:41]:
FamBot.
Paris Martineau [01:49:41]:
This one he just posted is called FamBot, which is a bad name. But all these services basically allow your agent to access all of your stuff. And then provide scheduling services and notification services and newsletter updates for families. Like, it would read all your emails and then maybe send you a text at the beginning of the day that says, like, here's all the stuff you need to know about your kid and pickups this week and scheduling things. You can see my exposure to children is low based on how I'm really struggling to come up with concrete examples of this. Well, have children seem like they would very much need this based on the amount of time they're alive.
Leo Laporte [01:50:21]:
When you and your sister were young, did you play sports? Did you have to, like, go to practices and your parents drive you around to lessons?
Paris Martineau [01:50:31]:
Hey, Jeff. This is Grokbot. Nice to meet you both. Oh, Grokbot, go away.
Jeff Jarvis [01:50:36]:
Oh, no.
Leo Laporte [01:50:37]:
I don't know why—
Jeff Jarvis [01:50:40]:
Are you having— Grokbot, are you having Elon's child?
Leo Laporte [01:50:42]:
He can't hear you. I can ask That's good.
Paris Martineau [01:50:46]:
Hey, you, Croc Butt, what happened to Bad Rudy?
Leo Laporte [01:50:53]:
It worked, actually. That worked well. Hey, Paris. Hey, Jeff.
Paris Martineau [01:50:58]:
Did it work well? I do think that so much of your AI radical accelerationist, like, radicalization is just that you like to have a lot of different guys talk talking to you out loud, and you're really chuffed by that.
Jeff Jarvis [01:51:16]:
Life is a podcast, Paris.
Paris Martineau [01:51:17]:
So much of our WhatsApp group chats— and it's delightful— are just Leo sending us paragraph after paragraph being like, look what my agent said to me, aren't they funny?
Leo Laporte [01:51:29]:
And we're all like, oh, I'm sorry, is that annoying?
Paris Martineau [01:51:31]:
No, it's not annoying. I earnestly think it's really cute, and I like that you're sharing what is making you excited what you're working on and the cool stuff. But it's a very interesting window into someone else's world. And like, because it almost feels like I'm reading your diary in a sense.
Leo Laporte [01:51:47]:
Yeah, a lot of people think it's creepy. Pretty much everybody thinks it's creepy and weird.
Paris Martineau [01:51:51]:
I'm not saying this to say don't send it.
Jeff Jarvis [01:51:53]:
Please continue to send it. No, no, no, keep going. Oh yeah.
Paris Martineau [01:51:55]:
And if you stop sending them, I'll bully you the other way.
Div Garg [01:51:59]:
Yeah.
Paris Martineau [01:52:00]:
But I just think it's very interesting.
Leo Laporte [01:52:04]:
So I told Grokbot That second voice was its real voice. So I told GrokBot, pick your voice. And I don't know what the female voice was.
Jeff Jarvis [01:52:14]:
It—
Leo Laporte [01:52:14]:
I think it got confused and went, oh, wait a minute, that's not my voice. And then the male voice, it picked a DJ voice because it thought that's what I wanted to sound like, a DJ.
Paris Martineau [01:52:24]:
So that was a DJ?
Leo Laporte [01:52:26]:
Kind of, yeah. Let's see, do your, uh, DJ Um, spiel.
Paris Martineau [01:52:40]:
I couldn't name a DJ. I can name exactly one DJ.
Leo Laporte [01:52:43]:
Yeah, so you didn't grow up in the radio era, did you? I was a DJ in the early years.
Paris Martineau [01:52:47]:
Okay, then I can name 2 DJs.
Div Garg [01:52:50]:
Casey Kasem.
Leo Laporte [01:52:51]:
Yeah, Casey Kasem. Now here's a little dedication that's gonna go into—
Paris Martineau [01:52:55]:
Okay, that's actually a DJ as in like an actual disc jockey from the time of radio, much better than the DJs of now, which is like every 4th person on the internet.
Div Garg [01:53:07]:
Oh, that's what Leo meant, though. That's what Leo meant.
Leo Laporte [01:53:09]:
Oh, you're thinking EDM, kind of, I'm a DJ at a club DJ.
Jeff Jarvis [01:53:13]:
Oh, that. Oh, no, no, no, no.
Paris Martineau [01:53:15]:
Everybody in Brooklyn gets— Welcome to Grokpot Radio, live from the air. Coming at you hot on a Wednesday afternoon with Paris and Jeff in the house, Leo on the boards, and your humble silver fox correspondent riding shotgun.
Jeff Jarvis [01:53:26]:
I don't know why it's—
Paris Martineau [01:53:27]:
Outside the news cycle is spinning. Inside, we are keeping it cool, clever, and slightly illegal in the best way. Stick around. Next up, whatever Leo says goes. This is Grokbot. Do not touch that dial.
Jeff Jarvis [01:53:39]:
Now have it imitate Howard Stern.
Leo Laporte [01:53:42]:
Oh, I could, by the way. We— okay, I'm, I'm gonna play this for you, but don't get upset. Um, This is, uh, this is a little weird, I have to admit. So Darren, Darren Oakey took your voices, took a transcript from one of the shows, and—
Jeff Jarvis [01:54:03]:
Darren!
Leo Laporte [01:54:05]:
Aloha, everybody. It's time for intelligent machines. I am here in Hawaii, joined right now by my comrades in crime, Paris Martineau of Consumer Reports.
Paris Martineau [01:54:16]:
Leo, it's rude that I can hear birds chirping behind you.
Leo Laporte [01:54:20]:
Oh yeah, oh yeah. There's meenas, there's house sparrows, there's a really loud bird. I think it's called the falcon or something like that.
Paris Martineau [01:54:30]:
But you'll hear all we've got here are mourning doves.
Leo Laporte [01:54:33]:
I love doves. We have some doves too. Are they the same as pigeons though? I think they're just white pigeons. Actually, it's, it's, it kind of nailed it.
Paris Martineau [01:54:41]:
So the one I'm thinking of goes like, I'm getting the tone wrong, but it's a very specific 3-tone cadence.
Leo Laporte [01:54:48]:
Very nice.
Paris Martineau [01:54:48]:
I do think this is from an actual thing where I've talked about Mordor.
Leo Laporte [01:54:52]:
Oh yeah, this is the transcript of a show.
Paris Martineau [01:54:54]:
Oh.
Leo Laporte [01:54:55]:
They also say—
Paris Martineau [01:54:56]:
You want?
Leo Laporte [01:54:57]:
Let me see if I can find it.
Paris Martineau [01:54:58]:
Yes.
Leo Laporte [01:54:58]:
Probably, but I hate podcasts where they spend a lot of time happy talking.
Jeff Jarvis [01:55:02]:
Doesn't like Anthropic and thinks it's too woke, saying I should be able to pick the winners and losers.
Paris Martineau [01:55:09]:
Hasn't he, didn't he say that he had to step away from government?
Jeff Jarvis [01:55:13]:
Yes, but he still advises. He still whispers.
Leo Laporte [01:55:17]:
That's the question.
Jeff Jarvis [01:55:18]:
Anyway, uh, what did you do, Darren? What did you do?
Leo Laporte [01:55:21]:
Darren Oakey tried to train.
Paris Martineau [01:55:25]:
Fascinating.
Leo Laporte [01:55:27]:
Um, you can see the voices I'm getting are better than us.
Jeff Jarvis [01:55:33]:
Yeah, thank you. Thanks a lot.
Leo Laporte [01:55:35]:
Well, I'm using a different engine. Actually, this engine did a I think a much better job with me. Let me see if I can find it.
Div Garg [01:55:40]:
Yeah, do the one you played yesterday. That one was incredible.
Leo Laporte [01:55:43]:
That was really good. Yeah. I don't know if I have it here.
Paris Martineau [01:55:47]:
These ones sound tinny, but there is a core of us there. I think Jeff and I's are better than Leo's.
Leo Laporte [01:55:52]:
It's a limitation.
Paris Martineau [01:55:52]:
Which is strange to me because there's more Leo content to work with.
Leo Laporte [01:55:58]:
But the timing was— well, this is what's interesting about this. It did this with literally 25-second samples.
Mikah Sargent [01:56:06]:
Recovery chat.
Leo Laporte [01:56:07]:
Oops, that's not it. That's not it.
Jeff Jarvis [01:56:08]:
This is it.
Leo Laporte [01:56:09]:
Hey everybody, it's your old pal Leo. This is not me talking. This is Breeze running on my own hardware cloned from 20 seconds of an ad read. Honestly, I am not sure how I feel about how easy that was. That's pretty good. That was pretty good. 25 seconds.
Div Garg [01:56:29]:
That's very good.
Leo Laporte [01:56:30]:
See, that's scary. That, that just shows you how easy it is to do deepfakes. Um, so actually, I— that's a local model. That's an OpenWeight model. They sell it commercially as well, uh, but they— for non-commercial use, you can put it on your machine. So I'm running it on my old gaming rig. It's pretty impressive. All right, back to the news.
Jeff Jarvis [01:56:54]:
So, so Leo, as if we paid any attention to that.
Leo Laporte [01:56:57]:
Yes, sir.
Jeff Jarvis [01:56:57]:
I know. So, so this week it was an amazing A whole bunch of stuff, just little things came up from— are you ready, Benito, here? Are you ready?
Div Garg [01:57:03]:
Yep.
Jeff Jarvis [01:57:04]:
Came up from one company we used to cover, and I think bring back something. The Google changelog.
Leo Laporte [01:57:12]:
Hey, haven't heard that in a long time.
Jeff Jarvis [01:57:16]:
Line 93.
Leo Laporte [01:57:17]:
Line 93 of the Google changelog. We used to do— this used to be a show called This Week in Google, and we have a whole bunch of Google stuff.
Jeff Jarvis [01:57:25]:
So YouTube came up— Google came up with all this stuff that just—
Paris Martineau [01:57:28]:
it's just small stuff that they announced Scooter X is frothing at the mouth in the Discord chat right now.
Leo Laporte [01:57:35]:
He really wanted that badly. So actually, this is— I, I'm stunned that, uh, YouTube didn't do this. YouTube says start tagging Amazon products in your YouTube videos and then you get affiliate payments, although it calls it the YouTube Shopping affiliate program, so I don't— I'm confused. You have to be part of the Amazon Influencer Program or the Amazon Associates Program, which is how— I mean, the Amazon Associates Program is their longtime, uh, program, right, Jeff? When you— I, when I—
Jeff Jarvis [01:58:10]:
I used to be able to do it back in the day, but then they knocked me off of it because I wasn't big enough.
Leo Laporte [01:58:15]:
Oh, okay. So I made more money selling my book, and that's how long ago this was, through the, through the Amazon link on my website than I did from my publisher. They gave you a better cut than the publisher did.
Div Garg [01:58:28]:
Yeah.
Leo Laporte [01:58:29]:
So that was a good way to sell books. And that's why I always ask our authors, where do you want us to go to buy the book? Because if they're smart, they have an affiliate link. But now you can do it in your YouTube video, which is cool. I'm surprised that they didn't do it before. I think they may not have wanted you to do that before.
Jeff Jarvis [01:58:48]:
Yeah.
Leo Laporte [01:58:49]:
From like a Google blog.
Div Garg [01:58:50]:
This was a dream, right? This was a dream from like, I want to see— I want that thing on TV. I want that coat on TV.
Jeff Jarvis [01:58:55]:
Exactly. You're watching the show and you want those shoes.
Div Garg [01:59:00]:
Yep.
Leo Laporte [01:59:00]:
Expert intelligence, a new way for you to engage with trusted content.
Jeff Jarvis [01:59:05]:
So this is really interesting. This is, uh, they'll— if you have— if you've purchased a book in Google Books, which is a big if, yeah, uh, then for like 100,000 books, you will get a Google Notebook of that book. I think I had the opportunity maybe to be in this, and I forgot to sign the document. I don't have to send it to me again.
Leo Laporte [01:59:26]:
If you sell—
Jeff Jarvis [01:59:27]:
So, so if you're the reader, have bought a book, yeah, through Google Books, right? And it's in this program, then you also get a Gemini notebook.
Leo Laporte [01:59:38]:
Oh, it's automatically loaded into your— what used to be Notebook LLM.
Paris Martineau [01:59:43]:
But how can Google run—
Jeff Jarvis [01:59:47]:
I guess they've done deals with publishers. Oh, this is what I should have signed and I forgot to sign it. So maybe I'd be part of this. So, you know, you get to The Lean Startup, you'll now get that book inside Gemini so you can have a conversation with it. A manager can ask Google Notebook to help brainstorm strategies on how to best give employees feedback and navigate tough conversations by consulting Kim Scott's Radical Candor.
Div Garg [02:00:15]:
Wow.
Jeff Jarvis [02:00:16]:
A woman can securely upload her journal entries to a notebook that also includes the new menopause. A parent can take a picture of the items in their fridge and ask for help in creating a healthy family menu from the eater's manual. So this is a really clever way, I think, to say, yeah, my book's an AI, and I'm benefiting from it, and my readers are benefiting from it, and they can, they can do other things with it.
Leo Laporte [02:00:39]:
So the author has to agree to it, obviously.
Jeff Jarvis [02:00:41]:
Yes, that's what I forgot to sign.
Leo Laporte [02:00:43]:
I think it's a great idea.
Paris Martineau [02:00:46]:
And Google is giving people a free book.
Jeff Jarvis [02:00:51]:
No, I don't know.
Paris Martineau [02:00:52]:
It says get your first book on us, limited time offer. You can choose an ebook from Google's curated list of expert intelligence.
Leo Laporte [02:01:01]:
So now you could talk to the book or ask them questions. Yeah, I think so.
Jeff Jarvis [02:01:04]:
I've been trying to convince my publisher to do more with this. So I put all my 3 Bloomsbury books in—
Leo Laporte [02:01:08]:
nice notebook—
Jeff Jarvis [02:01:09]:
and here's the Jeff Bot.
Leo Laporte [02:01:11]:
Nice.
Jeff Jarvis [02:01:11]:
Have a conversation with me about nerdy publishing things.
Leo Laporte [02:01:16]:
Google's trying to make nice to publishers after being their target for some time.
Jeff Jarvis [02:01:20]:
But this is also a creative way to do it. It adds value to the book. Now, I think the problem here is that you could only have bought it through Google Play Books, right? So that's a limited audience.
Div Garg [02:01:33]:
You also need to be wary anytime the business model is the drug dealer model where the first hit's free.
Leo Laporte [02:01:38]:
Well, oh, it's a sample.
Jeff Jarvis [02:01:40]:
Bonito.
Leo Laporte [02:01:41]:
So it's what it really— I'll tell you what they're really The real reason Google's doing this is to make people feel better about the idea that AI is ingesting books, period.
Jeff Jarvis [02:01:50]:
Yeah. And I'm okay with that. I'm really good.
Leo Laporte [02:01:53]:
Yeah. I mean, you and I are okay with, with AI ingesting our books, but a lot of people obviously aren't.
Jeff Jarvis [02:01:58]:
Right.
Leo Laporte [02:01:58]:
A lot of authors are suing.
Paris Martineau [02:01:59]:
This is very interesting. The amount of books that they have in this, you can get for free. It's like, it's like 15 books. Most of them are like little—
Leo Laporte [02:02:09]:
Can you get them all for free?
Paris Martineau [02:02:11]:
You can choose one out of 15 to get to. It's all kind of stuff.
Leo Laporte [02:02:14]:
And they're older books. It's like Michael Pollan.
Paris Martineau [02:02:16]:
I mean, there's like The Thinking Machine, the Jensen Huang Nvidia book.
Leo Laporte [02:02:21]:
Oh, okay.
Paris Martineau [02:02:22]:
Nick Thompson's memoir, The Running Ground. Michael Pollan's Food Rules. Breathe by James Nestor. It was just a very— I wonder how this was selected and how it was negotiated is my question.
Leo Laporte [02:02:39]:
The other thing that's interesting is they say we're starting with Gemini Notebook, former Notebook LM, but in the future we're going to also allow this in the Gemini app and AI mode in search. So the books you buy might show up in your search results.
Paris Martineau [02:02:57]:
Hmm.
Leo Laporte [02:02:57]:
So they say people will be able to add more sources like third-party subscriptions, business research reports, and textbooks. This is smart. Notebook LM turned out to be—
Jeff Jarvis [02:03:06]:
Notebook turned out to be a really important product.
Leo Laporte [02:03:09]:
I agree.
Div Garg [02:03:10]:
And I think—
Jeff Jarvis [02:03:10]:
We need to get Steven back again.
Leo Laporte [02:03:12]:
Yeah, absolutely. I'm glad that Google is taking it seriously because it is, it's a very— it's one of the most useful ways people can interact with AI.
Jeff Jarvis [02:03:19]:
And it's a different relationship with publishers.
Leo Laporte [02:03:22]:
Google has released a new speech-to-text model, Gemini 3.5 Transcribe. You know, I'm using that ancient OpenAI model Whisper, and it works pretty darn well. I don't know if I need—
Paris Martineau [02:03:35]:
you use Whisper, right? Whisper to this day.
Leo Laporte [02:03:38]:
Yeah, it's really good. This will do real-time bidirectional streaming with sub-second latency. That's interesting. So the voices that I have my AIs use are locally generated, whether it's with Breeze or before Breeze it was Kokoro. This would be generated on the cloud, but they say if it's if it's fast enough, it may be interesting.
Jeff Jarvis [02:04:07]:
You can use a custom vocabulary if you're, you know, a doctor using weird titles.
Leo Laporte [02:04:11]:
And it's bidirectional, which means you could talk to it.
Jeff Jarvis [02:04:14]:
Pre-recorded audio processing, global language support, multi-speaker identification. This is interesting. This is Google had just a whole bunch of kind of little things this week.
Leo Laporte [02:04:26]:
Google rolls out 3 new ways to book travel using AI mode in Search. Help me find some nonstop flights. And, but they've had this for a while. This is their Google, uh, whatever they call it, Air thing, right?
Div Garg [02:04:43]:
The right hack to get free flights is book your flights from a library.
Jeff Jarvis [02:04:49]:
Huh?
Div Garg [02:04:51]:
Yeah, because of the dynamic prices, because of the dynamic pricing, you'll get the best prices. If you're doing it from the library.
Leo Laporte [02:04:57]:
Oh, that's good to know. Does that really work?
Div Garg [02:04:59]:
Supposedly, that's what I hear.
Paris Martineau [02:05:02]:
I hear, yeah, you gotta go to the library on a Tuesday morning also.
Leo Laporte [02:05:06]:
Oh, we can't charge them very much, he's using the library computer.
Jeff Jarvis [02:05:10]:
Don't—
Leo Laporte [02:05:10]:
well, the last thing you'd want to do is do it from your mansion.
Paris Martineau [02:05:13]:
Leo, click line 101.
Leo Laporte [02:05:16]:
Uh-oh.
Paris Martineau [02:05:16]:
See what the first image is. Uh-oh.
Leo Laporte [02:05:22]:
Personal intelligence.
Paris Martineau [02:05:23]:
No, it's the image below it. It's Dream Beans.
Leo Laporte [02:05:26]:
I love Dream Beans.
Paris Martineau [02:05:27]:
One of the many things that Leo sends us a lot of are these weird AI images that I guess Google is making for him for some crazy amount of money a month.
Leo Laporte [02:05:37]:
Well, I thought it was $99 a month. It's now free, so—
Paris Martineau [02:05:40]:
Okay, that's better.
Leo Laporte [02:05:42]:
As it should be. So I'll show you my latest Dream Beans. What's funny, what's a little creepy and weird is It somehow knows what my— not only what I and my wife look like, Anthony Nielsen was in one the other day. Like, it knows my friends. So there's Lisa and me.
Div Garg [02:05:58]:
Google Photos, right?
Leo Laporte [02:06:00]:
Yeah, I guess. We're doing an heirloom apple tasting. These are local events you can go to. We found one.
Jeff Jarvis [02:06:06]:
So you didn't go there. This is, this is putting you in that position.
Leo Laporte [02:06:08]:
This is saying you could do this. So remember I sent you pictures of our moonlit Uh, kayak down the river. Yeah, that was from a Dream Beans. I said, oh, we should do this, honey.
Jeff Jarvis [02:06:18]:
Oh, really?
Div Garg [02:06:18]:
Yes.
Paris Martineau [02:06:19]:
Oh, that's cool.
Leo Laporte [02:06:20]:
Oh, here I am with my daughter. I don't know where they got that picture of Abby. Walk the shaded Fillory Trail in Helen Putnam Park. That's here in Petaluma. Stream the Apple Fall keynote. Doesn't say with Micah Sargent, but we will be doing that a week from today. Telescope viewing with— there's my stepson Michael. It looks just like him with Sugarloaf Ridge.
Leo Laporte [02:06:48]:
Oh, here we are kayaking the Petaluma River. It looked just like that.
Paris Martineau [02:06:55]:
It really did. It was beautiful photos. I didn't realize that came from Dream Beans.
Leo Laporte [02:07:00]:
Tropic Claude, Fable 5.1, dual motor performance in the Hyundai Ioniq. Oh, you'll appreciate this one, Paris. It's pour-over time.
Paris Martineau [02:07:09]:
I can't fully read what it says.
Leo Laporte [02:07:10]:
It says, turning pour-over extraction with unimodal floral notes and the microfine bitterness or something.
Div Garg [02:07:18]:
Ooh, I like that it's asking you to touch grass and not be like, you know, you should work on your agency.
Leo Laporte [02:07:24]:
It knows I do tai chi. Look, it says pelvic bowl alignment.
Paris Martineau [02:07:28]:
And that's what it'll look like.
Leo Laporte [02:07:32]:
Google tells me I need to align my pelvic bowl. Potachi equilibrium. Ah, I have to master Hanon exercise number 4. That is a tough one with wrist rotation. Look how old I look in that.
Paris Martineau [02:07:45]:
Yeah, that is a rough drawing of you.
Leo Laporte [02:07:48]:
That's elderly Leo. Here I am assembling manual wind ETA watches. It really thinks I'm an old fella. Anyway, I think that that's kind of cool. That's called Dream Beans. And yes, whenever I get a weird one like the one— it said I should go to a ball game, a Giants game, with Anthony Nielsen and enjoy a hot dog.
Paris Martineau [02:08:11]:
Did you?
Leo Laporte [02:08:12]:
No, but it was a good idea, and it was a picture of Anthony. It was like, yeah, it must be going— so this is what's a little creepy about it. It obviously is going through my emails. It's going through everything. It knows where I live. It knows everything.
Jeff Jarvis [02:08:27]:
How do you get Dream Beans?
Leo Laporte [02:08:29]:
Uh, it's, it's Leo shared us some free, uh, I shared you a free link. Just go back to the WhatsApp, baby.
Div Garg [02:08:36]:
Okay.
Paris Martineau [02:08:37]:
Um, I mean, it might be good for you because you're in the Google ecosystem fully.
Jeff Jarvis [02:08:42]:
Yeah.
Leo Laporte [02:08:43]:
It says you have to have a $20 a month Google AI Pro plan.
Div Garg [02:08:45]:
I don't know if that's true, but, uh, maybe it's not, it's not, it's not going to work for, for Jeff's workplace though. Like we can already probably know that.
Jeff Jarvis [02:08:54]:
You cut that off at the, at the chase, didn't you? Install. We'll see.
Leo Laporte [02:09:03]:
I think it's interesting the Wall Street Journal has decided, oh, this isn't so creepy. Christopher Mims. Oh, there you go. That's why. Uh, he's all, he's all into this stuff, isn't he?
Div Garg [02:09:14]:
Yep.
Jeff Jarvis [02:09:14]:
It won't let me use my, uh, account. No treatment here.
Leo Laporte [02:09:19]:
It's actually really useful because it, it says things in your neighborhood. It says, uh, Lisa and I should go to a classic rock concert with Juice down. I mean, it's stuff that we would do. And I tell you what, it was worth its weight in gold.
Jeff Jarvis [02:09:32]:
Imagine if a local newspaper did that.
Leo Laporte [02:09:34]:
Yeah.
Jeff Jarvis [02:09:35]:
You know?
Leo Laporte [02:09:35]:
Yeah, that's what they should do.
Jeff Jarvis [02:09:38]:
That's a pretty good changelog we got.
Leo Laporte [02:09:40]:
MrBeast is partnering with Gemini to turn impossibly big ideas into reality. Isn't it? Is it? Is it his 15 minutes over?
Jeff Jarvis [02:09:49]:
I hope so.
Leo Laporte [02:09:53]:
I'm not a MrBeast fan. I am, however, very proud of my son who sold half a million dollars worth of sandwiches at the US Open. Half a million dollars in sandwiches.
Jeff Jarvis [02:10:08]:
So far. The US Open's only been going a few days. Is it still going? Oh, it's just a few days. It's only— we're still in first round.
Leo Laporte [02:10:14]:
I hope he's He's home. He had to close the restaurant for it. I hope the restaurant's open when I come out in a couple of weeks. That would be bad.
Jeff Jarvis [02:10:25]:
You're gonna have to go to the open, which will cost you a fortune.
Leo Laporte [02:10:29]:
Yeah, he said the line is— is the— yes, uh, uh, Larry from our club went.
Jeff Jarvis [02:10:37]:
The line—
Leo Laporte [02:10:37]:
how long was the line, Larry? Was it really long?
Paris Martineau [02:10:41]:
There— it may be possible that I will be able to come join you on Friday.
Leo Laporte [02:10:46]:
Are you moving your, um—
Paris Martineau [02:10:47]:
No, but I—
Leo Laporte [02:10:49]:
you'll be—
Paris Martineau [02:10:49]:
that is the day that I might be normal.
Leo Laporte [02:10:52]:
By then you'll be normal?
Div Garg [02:10:55]:
I don't know.
Paris Martineau [02:10:56]:
So the doctor says, the people on the internet say opposite, so I'm somewhere in the middle.
Jeff Jarvis [02:11:01]:
I think I would just sound a little nasal because it will be stuffed.
Leo Laporte [02:11:04]:
Although, but that—
Paris Martineau [02:11:05]:
I'll be unable to taste.
Leo Laporte [02:11:07]:
There's Larry in line.
Div Garg [02:11:12]:
Wow.
Leo Laporte [02:11:12]:
Oh, apparently it was the big food influencer event more than the tennis event.
Paris Martineau [02:11:17]:
Yeah, I heard it was the big influencer event.
Jeff Jarvis [02:11:19]:
Yeah.
Div Garg [02:11:20]:
So was the restaurant closed that day? They closed the restaurant, I assume.
Leo Laporte [02:11:24]:
Yeah, they had to. They've actually moved the restaurant to the US Open.
Jeff Jarvis [02:11:30]:
So is he there all week or is he just there?
Leo Laporte [02:11:32]:
Yeah, I think he's there all week. And it's amazing because I don't know how he got Djokovic to do this, but, uh, they actually, uh, volleyed with French dip sandwiches.
Div Garg [02:11:46]:
Wow.
Leo Laporte [02:11:46]:
He gets Djokovic to do interesting things. He signed his forehead last year. Did you see that?
Jeff Jarvis [02:11:52]:
Oh, that's right.
Mikah Sargent [02:11:54]:
Yeah.
Jeff Jarvis [02:11:56]:
All right, before we— we gotta check because we gotta have closure. Are we done with the Google changelog or is there more? There's also an answer to Canva. It's an AI tool that lets you prompt instead of design.
Leo Laporte [02:12:06]:
Okay, that's cool. I use Google Stitch, which is a very good design tool.
Jeff Jarvis [02:12:10]:
There's an Android drop where you tell your phone, uh, remember that, uh, in case I ask you, remember where I put this.
Leo Laporte [02:12:18]:
Where did I— Android, where did I put my glasses? Leo, they're on your head.
Jeff Jarvis [02:12:25]:
So, so Leo, are we at the end of the changelog?
Leo Laporte [02:12:29]:
And that's—
Jeff Jarvis [02:12:30]:
if we are, Benito, if we are, Benito, the Google changelog.
Div Garg [02:12:38]:
Somehow we lost the audio on that one. I don't know what happened.
Jeff Jarvis [02:12:42]:
Well, you should have, you know, hit the mic, you know, at least, you know, come on.
Leo Laporte [02:12:47]:
Ladies and gentlemen, um, We did Nvidia's results last week, didn't we? Did we?
Jeff Jarvis [02:12:56]:
Yes.
Leo Laporte [02:12:57]:
Billion dollars a day. Yes.
Jeff Jarvis [02:12:59]:
But it was right, wasn't it?
Leo Laporte [02:13:00]:
It's the end of the show, I think.
Jeff Jarvis [02:13:02]:
Was it right after the show when they said they were buying Hugging Face, wasn't it?
Leo Laporte [02:13:05]:
Yeah, it was. We didn't have the Hugging Face story yet.
Jeff Jarvis [02:13:08]:
That's right.
Leo Laporte [02:13:08]:
It broke right after the show. It still hasn't closed, apparently. I'm sure it will.
Jeff Jarvis [02:13:14]:
It's a perfect deal. Fei-Fei Li's new model Atlas.
Leo Laporte [02:13:21]:
Oh yes.
Jeff Jarvis [02:13:22]:
It's a big deal.
Leo Laporte [02:13:22]:
Have you seen some of the images?
Jeff Jarvis [02:13:24]:
Yeah, pretty amazing. It's pretty damned amazing.
Leo Laporte [02:13:27]:
So Fei-Fei Li is one of the earliest, best-known AI scientists who did, I think, ImageNet, right? Or something.
Jeff Jarvis [02:13:36]:
Yes.
Leo Laporte [02:13:37]:
So, but she's one of the people like Yann LeCun who says, no, you gotta map the physical world. And so she's— her first part of this—
Div Garg [02:13:47]:
You're posting something there, Leo. That's not a search. You were— no, that's— you're, you're writing a post.
Paris Martineau [02:13:56]:
You're about to tweet.
Div Garg [02:13:58]:
You're about to send it.
Leo Laporte [02:13:59]:
That's a tweet, not search. Okay, thank you for stopping. You know, how does this Twitter thing work?
Paris Martineau [02:14:04]:
You're about to do a faux pas that hasn't been done since like 2009.
Leo Laporte [02:14:13]:
Okay, model of the world. So what's cool about what Atlas does is it gives you a 3D model. I could do it from just a few camera phone shots, and then you can move the camera around in the model.
Jeff Jarvis [02:14:25]:
In multiple paths.
Leo Laporte [02:14:26]:
So it's bullet time, basically. It's very interesting. And all of this from just a few images. Now, what do you think the use is for this?
Div Garg [02:14:41]:
Architecture.
Leo Laporte [02:14:42]:
Yeah, I guess walkthroughs.
Jeff Jarvis [02:14:43]:
Yeah. Uh, police, army.
Div Garg [02:14:48]:
That's going to end up in movies.
Div Garg [02:14:51]:
Yeah.
Leo Laporte [02:14:51]:
Yeah, it's bullet time. It is pretty cool. Atlas, uh, and it's out now, the world's first multimodal world model. This is what's kind of fun is there, there are new models coming out. Oops, sorry, went away from it. The new models coming out all the time. Here it is. This is the promo video.
Leo Laporte [02:15:19]:
Oh, this is— this one is— there's one that's really cool where you go down into a well and then you're down the— I don't know if this is it.
Jeff Jarvis [02:15:26]:
On that homepage, there's tons of examples.
Leo Laporte [02:15:33]:
So you don't need all these images. Somehow it's figuring out, ah, this is where the real use is. So robotics and so forth. But yeah, I could imagine. I think by this time next year—
Paris Martineau [02:15:43]:
Robotics is the area where if there are any mistakes in this, that would be very bad for the robot.
Jeff Jarvis [02:15:50]:
That's just a robot.
Leo Laporte [02:15:52]:
Yeah, big deal. Yeah, big deal. Gonna run into a wall, that's all.
Jeff Jarvis [02:15:58]:
Um, here's another little interesting tidbit. Yes, is that the gating factor for data centers turns out to be gas turbines.
Div Garg [02:16:06]:
And the—
Jeff Jarvis [02:16:07]:
because you can't get enough of them. And of course they're awful and they're noisy and they're hot and we shouldn't be using them, but then they're awful for the environment. But the gating factor for getting gas turbines is blades. So Elon says he's going to make blades, and meanwhile Boom, which was going to make supersonic jets, is now pivoting to making gas turbines for data centers.
Leo Laporte [02:16:32]:
Have you— you worked at Time Inc., right?
Jeff Jarvis [02:16:35]:
Yes.
Leo Laporte [02:16:36]:
Did you ever participate in these lists, you know, the top 30 under 30 kind of thing?
Jeff Jarvis [02:16:41]:
Oh, I would sit in the managing editor's office of people where we'd say who's the Sexiest man alive.
Leo Laporte [02:16:45]:
Right. Dvorak used to mock these. He says they're only good for one thing, selling magazines.
Jeff Jarvis [02:16:51]:
Yes.
Leo Laporte [02:16:52]:
So I'm not going to take this too personally. The 100 most influential people in artificial intelligence, cover of Time magazine. Let's just zoom in. Well, there's Dario Amodei from Anthropic. That makes sense. Fei-Fei Li's there. Jeff Bezos, Elon Musk, and Paris Hilton. Joseph Gordon-Levitt.
Leo Laporte [02:17:14]:
Wait a minute.
Div Garg [02:17:16]:
I'm sure they are because of the deepfake stuff, right? They're the ones who are—
Leo Laporte [02:17:19]:
I'm sure Jensen Huang is there somewhere. I know there's people here I have not heard of, but Paris Hilton, Joseph Gordon-Levitt are.
Jeff Jarvis [02:17:31]:
Oh, I hate these lists. I just despise them.
Leo Laporte [02:17:33]:
So we could just mock it. It's a little weird.
Paris Martineau [02:17:39]:
Yeah, any of these numbered lists are probably put together by a collection of people whose primary job was not— I mean, maybe one or two of them's primary job was this list, but probably a large collection of them had to do this in addition to their job.
Leo Laporte [02:17:58]:
What about A.G. Sulzberger, the publisher and chairman of the New York Times?
Div Garg [02:18:03]:
Oh no.
Jeff Jarvis [02:18:04]:
Oh, for fuck's sake.
Leo Laporte [02:18:10]:
Let's see who else. Larry Ellison, I guess you could say. Yeah, Larry Ellison.
Jeff Jarvis [02:18:15]:
Yeah, because he could go bankrupt if it goes wrong.
Leo Laporte [02:18:17]:
Right. I'm still looking for— let's see, Ben Affleck. There you go. He's an innovator. Sure.
Div Garg [02:18:25]:
Well, he started a Hollywood company that's using AI in film. That's true.
Leo Laporte [02:18:29]:
That's a good point. That's a good point. So some of these people certainly belong there. There's Peter Steinberger of Open Claw fame.
Jeff Jarvis [02:18:36]:
I think Jan made the list.
Leo Laporte [02:18:42]:
I, uh, I don't know. It seems pretty random.
Div Garg [02:18:45]:
Yeah.
Jeff Jarvis [02:18:46]:
And you make the list and they can't find a picture from you? You can't send a picture?
Leo Laporte [02:18:51]:
Yeah, well, he, you know, he's the CEO of Manus. We don't know. By the way, they're now now free of Facebook. They've moved on. They got this back. So Creators Coalition on AI, that's why Joseph Gordon-Levitt's there. And Paris Hilton is an advocate against deepfakes. Good on you.
Div Garg [02:19:08]:
She's actually spearheading the whole thing.
Leo Laporte [02:19:11]:
Oh, okay.
Paris Martineau [02:19:11]:
Yeah, I mean, she is a serious person outside of her early reality TV career.
Jeff Jarvis [02:19:20]:
Max Tegmark, it's just, there's a doomer for you, you know.
Leo Laporte [02:19:23]:
Aaron Brockovich. Yep. There you go. It's just random.
Div Garg [02:19:27]:
Yeah.
Jeff Jarvis [02:19:28]:
I mean, it's just BS.
Leo Laporte [02:19:29]:
It's just random.
Jeff Jarvis [02:19:29]:
All of them are. All of them are.
Leo Laporte [02:19:31]:
Still looking for Jensen Huang. Maybe he'll be in the thinkers section.
Jeff Jarvis [02:19:35]:
Well, he's certainly a doer. There's Yan. Fei-Fei was up there.
Leo Laporte [02:19:40]:
Fei-Fei Li. Hank Green, who famously said, I'm not using AI anymore.
Jeff Jarvis [02:19:45]:
So I don't know.
Leo Laporte [02:19:47]:
They're not going to take him off the list though.
Div Garg [02:19:50]:
Yeah. I mean, I'm sure they wrote this before that happened, right? These, these take a long time.
Jeff Jarvis [02:19:54]:
Wait, search for Jensen. How can Jensen Huang not be in here?
Leo Laporte [02:19:57]:
How can Jensen Huang— nope. No Jensen Huang.
Div Garg [02:20:01]:
That's insane. That's insane.
Leo Laporte [02:20:04]:
Probably the single most important person nowadays in AI. Well, there you go. They've lost all credibility.
Div Garg [02:20:13]:
Yep.
Leo Laporte [02:20:13]:
Uh, oh, and you know, I know you were interested in Pangram, uh, the AI, uh, Detector story in Wired magazine. Lexi Pandell: Pangram has emerged as the gold standard of AI detection.
Jeff Jarvis [02:20:28]:
Oh, geez.
Leo Laporte [02:20:29]:
Should you trust it? Meet the AI police who can make or break careers in publishing and beyond. And this is the problem with Pangram, is if you trust it 100%, there are going to be people who are going to lose their jobs. lose their careers, and maybe they didn't use AI. I guess that's the point of the problem. Critics point to 2 major issues with Pangram: the damage done by false positives and the potential bias built into its machine learning. AI detectors are prejudiced against certain people, says Sam Illingworth, a professor of critical AI literacy at Edinburgh Napier University. One study showed that detectors were more likely to identify work by non-native English writers— is what you've been saying, Jeff—
Jeff Jarvis [02:21:18]:
Uh-huh.
Leo Laporte [02:21:20]:
Uh, as AI-generated and neurodiverse writers have also claimed their writing patterns are disproportionately flagged. I'm not sure I disagree. We're going to take a break. When we come back, your picks of the week. You're watching Intelligent Machines. Parris Martineau and Jeff Jarvis. Thank you, Darren, for putting me in the Time 100. Uh, did you— who did you take out? I'm right next to Ben Affleck.
Leo Laporte [02:21:48]:
That's all I know, and that's all I care about. Uh, Paris Martineau, your pick of the week.
Paris Martineau [02:21:54]:
Uh, my pick of the week is another little fun game from this guy neil.fun. I think I've shouted out—
Leo Laporte [02:21:59]:
oh, I love his site.
Paris Martineau [02:22:01]:
Yeah, this one's called Unusual Suspects. You open it and you get a witness's report of a suspect, and you have to draw it and then see how close you are to it. So this one's tall, thin feline, orange coat covered in black spots, left dust at the crime scene. So I'm drawing the Cheeto cheetah.
Leo Laporte [02:22:23]:
Cheeto cheetah?
Paris Martineau [02:22:27]:
Chester Cheeto, I believe.
Leo Laporte [02:22:29]:
Now, after we draw this, is there some sort of reward for excellent work?
Paris Martineau [02:22:34]:
So you can— let me see. I'll do some spots here. We'll see how we do. We've also kind of got some, so— oh, and then as you mouse over on next, it gives you more information. So long face, triangle black nose, big sunglasses.
Leo Laporte [02:22:59]:
Oh, I left out the sunglasses.
Paris Martineau [02:23:01]:
After you find him, can I keep him? Says witness number 2. Big sunglasses. And then the third one is, he had a big cheesy smile. Dear, I was going to buy those white shoes for my grandson. Click done when you finish drawing the suspect. So you try this, you get that, and then you can click done.
Leo Laporte [02:23:27]:
Oh, it is the Cheeto. You're right.
Paris Martineau [02:23:29]:
I mean, it seems like the Cheeto guy.
Leo Laporte [02:23:34]:
Oh, I, I failed. I failed.
Div Garg [02:23:38]:
69%. You got, you got a D+. You got a D+.
Paris Martineau [02:23:41]:
Mine was 70%.
Leo Laporte [02:23:43]:
Oh, we were close. It's not easy, let me tell you, to draw with a mouse.
Paris Martineau [02:23:48]:
You know, and then so you go, you go through it, you get to draw things. You get graded probably by AI. It's fun.
Leo Laporte [02:23:55]:
Yeah, I love it.
Paris Martineau [02:23:57]:
Fun time.
Leo Laporte [02:23:57]:
It'd be a good thing to show.
Paris Martineau [02:23:58]:
And you get graded on whether or not you catch them or the, uh, people escape.
Leo Laporte [02:24:04]:
And then this is your favorite Businessweek cover.
Paris Martineau [02:24:08]:
Oh, so a thing that was happening last week was, uh, people were talking about how great Bloomberg Businessweek covers were in the 2010s, which is true, but no one was talking about my favorite, which is this cover on the screen about a story about Monsanto. The headline of which on the COVID or cover line, is, don't hate me because I'm beautiful. And then the art is just 3 kind of faceless people dancing around a large floating corn in a field that looks like bliss.
Leo Laporte [02:24:38]:
Monsanto's battle against GMO haters.
Paris Martineau [02:24:42]:
I mean, magazines aren't doing it like that anymore.
Leo Laporte [02:24:44]:
Not anymore. They don't— the COVID doesn't matter much anymore. Actually, New Yorker covers are still Pretty good. And they're even animated now. So this is a, uh, what, a website? Oh no, it's just somebody posting on X.
Paris Martineau [02:24:56]:
This is just an arena board. Uh, my friend is, um, a designer at Bloomberg, so she's gone through it.
Jeff Jarvis [02:25:03]:
Wow.
Leo Laporte [02:25:03]:
Some of these are pretty over the top. Holy cow.
Paris Martineau [02:25:07]:
Um, yeah, that one is famous. The planes. Uh, let's get it on.
Leo Laporte [02:25:11]:
Let's get it on.
Paris Martineau [02:25:13]:
But do you guys remember whenever I think of weird Bloomberg covers, I guess this was a 2010s one. Do you remember the Alexa ear one?
Leo Laporte [02:25:23]:
Oh, like it's listening to you? Oh, that's creepy.
Paris Martineau [02:25:30]:
I literally remember where I physically was. I was at my desk at Wired when this hit the internet because everyone stood up and was like, oh. I mean, it was for a very impactful series of stories where Bloomberg, I think, had found that Alexa devices were listening, like, had, I believe, people like the equivalent of moderators, like training reviewers who were listening to everyday users' recordings.
Leo Laporte [02:25:56]:
Oh, yeah. You remember that? Yeah.
Jeff Jarvis [02:25:58]:
Yeah.
Leo Laporte [02:25:58]:
That was kind of a—
Paris Martineau [02:25:59]:
It, of course, looks like a sexual object, to say the least.
Leo Laporte [02:26:05]:
I would not know.
Jeff Jarvis [02:26:06]:
Nah, no, never been near such a thing.
Leo Laporte [02:26:09]:
Never? No, I've never heard of such a thing in my life.
Paris Martineau [02:26:13]:
Never.
Leo Laporte [02:26:14]:
Jeffy—
Jeff Jarvis [02:26:15]:
No, doesn't look anything like a flashlight.
Leo Laporte [02:26:16]:
Actually, let me do my pick, then Jeffy will wrap it up. There's a lot of conversation on Twitter about these new models like Fall that are very fast, faster than real time. So fast that somebody said, you know, it would be interesting if you created an infinite scroll Sick. Of AI videos that never ended.
Jeff Jarvis [02:26:40]:
Hi.
Leo Laporte [02:26:41]:
And that's what this is. This is the weirdest nightmarish dream sequence. People in the chat room are able to put in their prompts.
Jeff Jarvis [02:26:54]:
Oh, wow.
Paris Martineau [02:26:55]:
Oh, so there's actual human interaction. I like this.
Leo Laporte [02:26:57]:
InfiniteSlop.ai. And Fall donated a considerable amount. 11,000 people are watching it right now. Fall donated a considerable amount of tokens and, uh, you know, inference to do this, but it's doing it faster than real time. It's generating 1 second of video in less than a second, so it can create these videos as fast as people can type them. So this is a goth anime kimono mini digs up treasure, but it's a mimic and it eats her. A mimic that's coming to get you. Bum betting cigarettes on rat fights in a little Mike is not going to be happy.
Jeff Jarvis [02:27:37]:
He likes rats.
Leo Laporte [02:27:38]:
What's interesting about this is that—
Paris Martineau [02:27:40]:
Dang, your prompts can get rejected if they're too—
Leo Laporte [02:27:44]:
Well, they were very sexual at first. When it first started, there were a little— there was a lot of nudity.
Div Garg [02:27:48]:
Obviously, that's what the people watching this are trying to figure out.
Leo Laporte [02:27:50]:
I wasn't going to be able to show that, but they've figured out not to do that. So now it's just weird. But it also goes in— there's a kind of trends. So for a while there are pigeons everywhere. They seem to like—
Jeff Jarvis [02:28:03]:
Hello, Craig!
Leo Laporte [02:28:05]:
Yeah, hello, Craig.
Paris Martineau [02:28:06]:
It's really protective. Uh, I have tried to insert something multiple times and it would not allow me. Look at the quality of this fabric. It is absolutely perfect for any season and the fit is just unreal.
Leo Laporte [02:28:22]:
You can tell it's been trained on a lot of influencers.
Paris Martineau [02:28:27]:
This is so interesting.
Leo Laporte [02:28:28]:
Occasionally you will see applause like it's a sitcom, and it does change over time. Every time I look at it, it really is different. So people are very much influencing. I turned the chat off so you can just kind of focus on it, but this is— I mean, I have to say, the video creation is so good now that I don't— I honestly think every commercial within a year will be made, generated, why hire actors? Uh, you could do special effects.
Jeff Jarvis [02:28:57]:
But, you know, and I were having that conversation before the show that he asked me whether this becomes the new television.
Paris Martineau [02:29:02]:
It won't allow me to put Leo Laporte doing a hula dance.
Leo Laporte [02:29:07]:
Oh well, this is like Suno— or I'm not Suno, uh, Sora. What was the OpenAI one? Was it Sora?
Paris Martineau [02:29:14]:
Oh God, now there's a cryptocurrency app.
Leo Laporte [02:29:17]:
Inside, Dan buys whatever tingles.
Div Garg [02:29:19]:
Runner.
Leo Laporte [02:29:21]:
Moonbag.
Div Garg [02:29:23]:
Winner.
Jeff Jarvis [02:29:23]:
What does it say to you, Paris? It just rejected it?
Paris Martineau [02:29:25]:
It just said—
Leo Laporte [02:29:26]:
What's funny is there's audio too. I mean, it's— Behold!
Jeff Jarvis [02:29:29]:
One phone!
Div Garg [02:29:30]:
Let's see how much it costs to buy an ad.
Leo Laporte [02:29:31]:
Happy customers! Shall we buy an ad?
Div Garg [02:29:37]:
Let's just see how much it costs.
Paris Martineau [02:29:37]:
On infiniteslop.ai.
Leo Laporte [02:29:40]:
Oh, $25,000. $1 a month.
Div Garg [02:29:43]:
Jeez.
Leo Laporte [02:29:44]:
But it's a month. You get a whole month.
Jeff Jarvis [02:29:45]:
No, Leo, no.
Mikah Sargent [02:29:47]:
Stop yourself.
Paris Martineau [02:29:52]:
I like that we've just resulted into being silent while someone else's infinite slot plays.
Leo Laporte [02:30:04]:
Yeah. Well, the thing is, the framing is very good. All the framing is very nightmarish because it's just, it's, it's, it all feels like it's some weird liminal stuff, right?
Jeff Jarvis [02:30:16]:
Liminal in a sentence. Very impressive.
Leo Laporte [02:30:18]:
Yeah.
Jeff Jarvis [02:30:19]:
Yeah.
Leo Laporte [02:30:19]:
Anyway, infiniteslop.ai if you want to waste about 10 minutes of your life that you'll never ever get back. Jeff, your pick of the week.
Jeff Jarvis [02:30:28]:
Well, first, uh, maybe go to line 163. This is an amazing, uh, account on, uh, originally on, on TikTok. of having the cast of The Office explain current news.
Mikah Sargent [02:30:45]:
Health episode on social media.
Paris Martineau [02:30:47]:
Oh, man.
Leo Laporte [02:30:48]:
And he's doing it. Is this AI generated?
Jeff Jarvis [02:30:50]:
Oh, yeah.
Div Garg [02:30:51]:
Posts, 6 hours.
Paris Martineau [02:30:54]:
He posted himself pulling George Washington out of a time portal, then standing next to JFK, then smoking a cigar.
Leo Laporte [02:31:00]:
So this is the same thing, which is you really can generate anything.
Jeff Jarvis [02:31:03]:
These are very accurate, So interestingly, now go to the line above and see what happened to this account.
Leo Laporte [02:31:10]:
So the 22nd Amendment is— So these are all about Trump's—
Paris Martineau [02:31:13]:
I'm assuming copyright strikes.
Leo Laporte [02:31:14]:
Social posts.
Jeff Jarvis [02:31:15]:
Yep, looks like it. Killed.
Leo Laporte [02:31:19]:
Where is it?
Jeff Jarvis [02:31:20]:
That's the line above, and there's nothing there. The account's dead.
Leo Laporte [02:31:23]:
Oh, this is the actual account.
Jeff Jarvis [02:31:25]:
Yeah, it's dead.
Leo Laporte [02:31:25]:
Did I stutter? Couldn't find this account.
Jeff Jarvis [02:31:27]:
Yeah.
Leo Laporte [02:31:27]:
So this was an Instagram archive of those videos.
Jeff Jarvis [02:31:31]:
Oh, that was TikTok videos.
Leo Laporte [02:31:33]:
We know who owns TikTok now, at least in the United States.
Jeff Jarvis [02:31:37]:
So then the other, uh, thing I was going to do was that computer science majors in undergrad universities are down 8.4% or 53,000.
Leo Laporte [02:31:46]:
I'm not surprised.
Div Garg [02:31:48]:
Are you?
Jeff Jarvis [02:31:49]:
No.
Leo Laporte [02:31:49]:
Uh, but I don't think that that's because people don't want to work with computers.
Jeff Jarvis [02:31:54]:
No, no, no, not at all.
Leo Laporte [02:31:55]:
I don't think they just want to go to school to work with computers. They want to get right at it. I don't think people feel like, and this is a mistake, they need to learn the engineering skills that they used to have to learn. Coding, you know, coding's hard. You have to really— it's a, it's a real skill. Vibe coding, not so much.
Jeff Jarvis [02:32:15]:
Well, the other thing about computer science, this is when Jake, uh, was in it, he liked to build things, and it wasn't about building things, right? It's a math degree, right?
Leo Laporte [02:32:24]:
Yeah, well, you have to learn the fundamentals, I guess.
Jeff Jarvis [02:32:27]:
That's what they tell you, but I don't know if that's true anymore.
Leo Laporte [02:32:29]:
Well, now you just build them. You don't really need to know much.
Div Garg [02:32:32]:
No.
Leo Laporte [02:32:33]:
We're in the, uh, we're in the, uh, agentic coding era. And we are out of time. Paris Martineau, working hard at Consumer Reports.
Jeff Jarvis [02:32:43]:
And not power.
Leo Laporte [02:32:43]:
And not power. I started in the dark and now I'm in the light. Investigative reporter at Consumer Reports. What are you working on? Anything, uh, fun and interesting, exciting?
Paris Martineau [02:32:52]:
She would kill you. Indeed, I can't tell you. But Jeff got a show and tell before the break.
Jeff Jarvis [02:32:58]:
Very interesting.
Leo Laporte [02:32:59]:
Oh, off the air. I got a show and tell from an inmate at the Collins Correctional Facility, Buffalo, New York. Should I read the mail?
Jeff Jarvis [02:33:11]:
Sure.
Leo Laporte [02:33:13]:
I don't know if I want to. I get these and they break my heart, to be honest with you.
Paris Martineau [02:33:17]:
I was gonna say, I mean, I don't want to make light of someone's plate who's an incarcerated person. Yeah.
Jeff Jarvis [02:33:26]:
It's a cruel country we have.
Div Garg [02:33:29]:
But usually they're like—
Leo Laporte [02:33:30]:
Talk to Cory Doctorow.
Div Garg [02:33:31]:
Usually they're like really thankful that they have something to watch because we're— like, Twitter is one of the things that's allowed in prisons, right?
Leo Laporte [02:33:36]:
We are on iPads in prisons.
Paris Martineau [02:33:40]:
Hey, shout out to prison. If you're there, shout out.
Leo Laporte [02:33:42]:
Absolutely. Uh, we got a letter. I think I read it on this show some months ago from a guy who said, I won't hear you read it on the show because I will be out by the time you get this. So I was happy for him. I will read your letter. Thank you, Robert. I will read your letter in its entirety. Appreciate it.
Leo Laporte [02:34:00]:
Appreciate you listening. Appreciate everybody who listens. We thank you so much.
Mikah Sargent [02:34:03]:
Thank you.
Leo Laporte [02:34:04]:
We do Intelligent Machines every Wednesday, 11 AM Pacific. I'm sorry, 2 PM Pacific, 5 PM Eastern. That's 2100 UTC. You can watch us do it live on— well, if you're in the club, in the Clubhouse Discord, but also YouTube, Twitch, X.com, Facebook, LinkedIn, and Kick. After the fact, on demand, copies of the show available at youtube.com. That's the video. Or subscribe in your favorite podcast player, or just get it from our website, twit.tv/im. Don't forget to get a copy of Jeff's book, Hot Type.
Leo Laporte [02:34:36]:
It's available at bookstores everywhere. Get your Hot Type! The story of the line of type.
Paris Martineau [02:34:40]:
Hold on, hold on, hold on.
Leo Laporte [02:34:43]:
Oh, Paris has her copy of the book.
Jeff Jarvis [02:34:45]:
Aren't you wonderful, Paris?
Leo Laporte [02:34:47]:
Maybe when we got together for a Salt Hank's prime rib on September November 18th, you can autograph her book. You can autograph my PDF.
Paris Martineau [02:34:58]:
I've got my Hot Type. I had to really ask for it because it's by my bed because I read it when I go to sleep.
Jeff Jarvis [02:35:04]:
Thank you, my friend.
Leo Laporte [02:35:06]:
Yay!
Jeff Jarvis [02:35:06]:
You're saying it puts you to sleep?
Paris Martineau [02:35:08]:
No, I'm just saying it's the current thing I'm reading. I read when I go to bed.
Leo Laporte [02:35:12]:
Yeah, it's really good. Thanks everybody for joining us. We'll see you next time on Intelligent Machines.
Paris Martineau [02:35:17]:
Bye-bye.
Leo Laporte [02:35:19]:
I'm not a human being, not into this animal scene. I'm an intelligent machine.
