OpenAI Says AGI Is Here

00:00:00:07 - 00:00:24:00
Unknown
Coming up next, Jeff Jarvis and I dig into OpenAI, declaring they've achieved AGI with their new Astra model. Feels a little like marketing to me. I don't know about you, but also a little bit more information on the swarm of agents that hijacked a German wiki with thousands of fake pages that came from OpenAI as well. Some interesting things to talk about.

00:00:24:00 - 00:00:55:03
Unknown
Their Google escaping an ad tech breakup, and how that relates to artificial intelligence. Hikers getting rescued from Mount Shasta after trusting Gemini to plan their trip. The short answer is don't do that. And we take a look at food in the in the view of generative AI and why it's so not at all appetizing. Just a quick note to restaurants don't use AI for your menus, please.

00:00:55:03 - 00:01:10:25
Unknown
That's all. Coming up next on this episode of the AI Inside podcast.

00:01:10:27 - 00:01:31:01
Unknown
What's going on, everybody? Welcome to another episode of AI Inside the Show, where we take a look at the AI that is layered throughout the world, the technology with one of two of your hosts on screen, right? Well, actually 1.5, because my dog Bronson just showed up, as you can see his tail wagging back and forth. He's hopefully not going to talk very much on this podcast, but you never know.

00:01:31:01 - 00:01:46:16
Unknown
He's looking kind of needy right now. I'm one of your host, Jason Howell. On the other side of the country is Jeff Jarvis. How are you doing? Towel tails wagging. Yes. That was just me. You know, you know, a little too much now.

00:01:46:19 - 00:02:05:17
Unknown
Yeah. How you doing? All right. How are you doing? All right. Let everyone know. Yes. Clearly we are a day early. We are here. My fall of going to status news event in Manhattan tomorrow. So. Okay. Big old, big old media, by the way, last if I may, real quickly, last week on the show, I said, hey, you could get an autographed copy of the book.

00:02:05:19 - 00:02:24:08
Unknown
And then Jason went to the Lincoln and said, no. What? So I dropped off another five copies. Those are sold out. I'm going back Friday, Simon. But if you go to Jeff Jarvis com, there's a link to Montclair Book Center. You should be able to click on it, I think. But if you don't just come back a few days later because there's more books coming in and I'm heading out there.

00:02:24:09 - 00:02:41:29
Unknown
Nope. Do the same thing. Yeah. I mean, the link works. It takes you to the site, it takes you to that. You're trying to fix it. Oh, yeah. Yeah yeah yeah. Oh, yeah yeah, yeah, just check back. Check back. Probably Friday afternoon if you, if you're looking. I'm delighted the people that bought it. But we've lost some sales which hurts too.

00:02:42:01 - 00:03:01:01
Unknown
So anyway so I apologize folks. We know what's going on. We're trying to fix it together. So yeah, we'll get there eventually. You'll get your signed copy from Mr. Jeff Jarvis if you'd like. Of of hot type, if that is the thing you're looking for. Right. As we were about to go live, I looked on the floor and I realized I have a pair of Gemini socks.

00:03:01:02 - 00:03:20:28
Unknown
This is a very random thing to show off on a podcast, but apparently I have a pair of Gemini socks. I should wear these. They're very nice. Or you could give them away to a nice user. You know, we could keep your feet comfy. I mean, they're actually really nice socks. Well, they're keeping it. We'll see, we'll see.

00:03:21:01 - 00:03:41:02
Unknown
Why don't we talk about some news? Today's rundown is a little a little bit different. Not in the beginning. We've got some big stories talk about before the break. And then afterwards, we just got some fun stuff to talk about. Because being one day early, you know, and it changes kind of how much news comes through. And then also I think maybe the Apple an event this week is kind of sucks.

00:03:41:08 - 00:04:03:13
Unknown
Stealing the yeah stealing the oxygen a little bit on big news and stuff. But but last week I don't know how much how much you you cared. How like, do you use ChatGPT? Very much. Not much. I've got to decide now whether to subscribe, because Astra is supposedly really, really, really, really good. And it's available at the Buck plan.

00:04:03:13 - 00:04:26:04
Unknown
Right? That it will be. I still don't know that I have it like I know. Oh, maybe, maybe not though maybe maybe or maybe I do. I don't know, it was a it was a crazy weekend. I was nowhere near my normal technology life throughout the weekend because my daughter was being recruited for swim. And so I was on a trip with her to Arizona, being wined and dined.

00:04:26:04 - 00:05:04:02
Unknown
It was a really cool experience, I have to say. I'm very proud of my daughter for for having a school be so excited to to bring her to it. But anyways, I was very like not connected very much to the technology news over the weekend. And so I know that GPT six Astra is announced and is here, but my last understanding is that it was only hitting certain, you know, certain people like Daybreak Access program coming to a limited group of organizations, that sort of stuff, but eventually rolling out to plus pro-business enterprise API and AWS in the coming days.

00:05:04:03 - 00:05:20:01
Unknown
I don't know that I have access to it yet. Anyone who's watching the live stream in the chat, if you have access to it, let me know. But yeah, I mean, don't they all say it's the latest and greatest and you got to get it because it's the best thing and blah blah blah. This one got a lot of big publicity.

00:05:20:01 - 00:05:42:03
Unknown
This one got a lot. I mean, you know even so Jensen so so the company is will discuss said AGI is here. Yes Jensen Wong said that you know at two Jensen. Really? Yeah. So Daniel confuse it. Okay. Daniel, give us a quick review. What do you think of it? Yeah, totally. Well, while we talk about this, throw a throw a little chat in the in.

00:05:42:03 - 00:06:09:20
Unknown
We'll feature it and let us know like what you've noticed. And so basically I have to get in there and I have to use it. Yes. So you mentioned the whole AGI thing. Greg Brockman from OpenAI basically went ahead and said, it's not unreasonable to feel that we are now in the AGI era. It's not unreasonable not going to say AGI is here, but I'm just saying I'm just saying if you want to say it, I won't make fun of you.

00:06:09:21 - 00:06:28:23
Unknown
It's basically what he's saying. Also, it just makes you wonder what like what? What the heck does that even mean? I don't even know. Yeah, it's it's so frustrating. And I've said it again and again and again on the show, so I don't want to go on forever. But there is no definition of AGI. Plus, I think this notion of going after general intelligence is a mistake.

00:06:28:23 - 00:06:51:09
Unknown
I think specialized models and specialized uses of them are where the action is going to be and the value is going to be, and it's just so much breastfeeding male geek hubris. And it doesn't serve any function. In fact, I think it it causes more damage because people say, oh, I'm scared. We've been to regulate the heck out of this, and I think it's going to have an impact.

00:06:51:09 - 00:07:13:14
Unknown
So AGI in my view, remains and will remain a myth. Yeah. As you mentioned, Jensen Long also saying AGI has arrived from ChatGPT to zero one to Astra in four years. Can you believe it? Can you believe how far we've come in such a short period of time? I mean, it is impressive. It is impressive. It is amazing.

00:07:13:14 - 00:07:37:07
Unknown
It's wonderful. But yeah, just leave it at that. Yeah. Let's go to this extra step. Oh, it's such a marketing term. It's just, it's it's so marketing driven and doesn't really mean anything. And it's marketing toward investors, not toward users to the user. Can it do what I wanted to do or not? Right. That's the question. Well yeah.

00:07:37:08 - 00:07:59:26
Unknown
Yeah. Yeah. Right. Open AI still pre IPO. We still don't have the the IPO announcement. So of course you know they they all want to be able to say whatever. And again as you just mentioned there is no real core definition of what AGI actually means. Or do you know do you know AI when you're looking. It's like porn.

00:07:59:27 - 00:08:22:20
Unknown
Do you know it when you see it like I know it when I see it? And, you know, is AGI kind of in that weird, like, an undefinable area? Do I doubt that it'll happen eventually, whatever it may be? I mean, I suppose to some of the definitions, I suppose so, but I think right now for these companies, it's, it's really a marketing term.

00:08:22:20 - 00:08:48:04
Unknown
And yeah, I think, I think you're spot on. It's, it's more about getting investors really excited about this company being the one that gets there because everybody understands the term loosely, but nobody they know the chances exist. Right. Exactly. They know it exists, but they don't really understand what it actually means. And this notion of, of general that it can do anything anyone can do.

00:08:48:05 - 00:09:07:21
Unknown
As as the oft quoted here, Yann LeCun, says, nobody, no human being is general. You can be good at chess and you can be crappy at writing, good at music, and crappy at design. You know, no one's general. It's a ridiculous goal.

00:09:07:23 - 00:09:49:23
Unknown
Yeah. Who is it? AI researcher Ethan Mullick had a had a description for AGI said AGI is not a well-defined, termed, well-defined term, which we clearly agree. Said that what we have is jagged AGI. Yes. Better. Better than average humans in many areas. Worse in others. Yeah. I was listening to a podcast by Cal Newport. He was very smart, who was a mathematician and computer theoretical computer scientist, and he was discussing the.

00:09:49:26 - 00:10:12:25
Unknown
Civilizations post and also discussing where these advances are. And he said that what's happened is that he was talking about spiking this. And we know that these models are really good at coding point one. We now see that against certain but not all of the major math tests, they're going to be pretty good. Yes. Well, there was just news about this today, right.

00:10:12:27 - 00:10:44:29
Unknown
That I don't know that we even had in in the rundown. But anyways, continue and I'll dig that up. Right. So so so just because it can reach those spikes doesn't mean that it fills in everything else. Yes, totally. It doesn't. And so again, it's just a useless discussion. It doesn't help. The other thing that that Newport said is part of the problem is, is they're trying to give the impression of generality.

00:10:45:02 - 00:11:01:10
Unknown
But look, if it can do all this and it can do all that, it must be able to do everything, must be able to do everything. Yes, but it can't. It doesn't. And again, I think it skews the design process. I think it's really just kind of a mess all around.

00:11:01:13 - 00:11:29:01
Unknown
So yeah. So but, but everybody seems to think that that Astro is great and I gotta, I gotta play around with it on my own. You know, OpenAI saying it's the best one for computer use and all the things I. So you subscribe to Claude now? I mean, I do have a subscription to OpenAI or to ChatGPT. I do have, yeah, subscription still to Claude.

00:11:29:02 - 00:11:54:01
Unknown
It's it's the one that I continue, you know, I've built so much of like my current process and everything inside of that. And granted, it is portable. Like I have a whole, you know, on disk structure, like files and folders and everything, and it feeds into itself and it's self improving. And it's all these things that I continue to work with and work inside of that, that framework essentially.

00:11:54:01 - 00:12:34:21
Unknown
So, you know, technically, every time I use it, it gets better and smarter than the time I used it last for everything that has to do with my business. So that's great. And it's it's technically portable because it is a folder and file structure with markdown files and folders for different workstations. Let's say it's, you know, it's kind of a complex layout, but makes sense if you read the MD files that I've created and as agents would do, as these LMS would do, so technically I should be able to point it at or point ChatGPT and potentially, I guess, this Astro model at it and pick up where I left off.

00:12:34:21 - 00:12:50:10
Unknown
I don't know, I think technically that should be pretty easy, but I just I keep not doing it because I, because my, my process is over there and I already know how to do that. And I don't want to start from square one again. Right. It's like switching from Android to Apple at this point. This is where I think.

00:12:50:11 - 00:13:13:20
Unknown
But I think this is where harnesses come in. Yeah. And that's kind of a later stage of all this. Okay. Explain. Harness. Yeah. Well, just a harness just controls. I'm not going to go to explain it, but but a harness is something you can have. I wouldn't say locally, but it's yours that can then control multiple models. Oh I see okay.

00:13:13:21 - 00:13:50:05
Unknown
Right. Okay. So this is where I think to some extent not really true but hugging face. And what's the one that stripe bought eventually May comes I think that there may be a layer of harnesses. I keep wanting to call them saddles. Harnesses that allow you to slip in different models, or also use it as an orchestrator to say what's best for this model versus that model versus this model versus that model, so that you're the business intelligence of your own stuff, stays resident with you, and the model becomes commodified.

00:13:50:11 - 00:14:11:02
Unknown
I think that's where we end up going. It's not a great sign for the model makers because they want to keep you. They want to keep you in their thing. Right? But if you have all those files and you can point it to Astra and say Astra, here's where I was with Claude. Beat him. Right. Yeah. So Daniel says harness is the app.

00:14:11:04 - 00:14:36:17
Unknown
Hermes is a great harness. Open clause one. And Pegasus says harnesses like open code, py, etc.. Okay, so harness is the thing that directs the traffic to the different the different models essentially that you're using. I think that's a little oversimplified. That's my fault for oversimplifying. But yeah, I think that it it's a, it's a model that allows you to have a control layer that work.

00:14:36:21 - 00:14:49:26
Unknown
Okay, okay. And so what I wonder is. Whether there starts to be a marketplace of harnesses.

00:14:49:28 - 00:15:10:26
Unknown
Like an app store says, this is really this is this is already thought through for math problems or this is thought through for business problems, or this is thought through for medicine and so on and so forth. Right. And the generality of the model underneath still becomes useful, but there's specialization above it. And we start to get a stack here.

00:15:10:28 - 00:15:33:08
Unknown
But I speak out of ignorance, Daniel, than I do. Clearly he does. He says it's like the cloud app you chat in, but not restricted to just Claude. And I mean, that makes that makes a lot of sense to me. He also said I have Claude ChatGPT and a few external APIs working on one file, which, yeah, that's that's great.

00:15:33:08 - 00:15:53:14
Unknown
That's probably really powerful. I'd be worried that one thing would be making a change to that one file at the same time as another thing, and then let them get out. Yeah. May the best file win. Or maybe the best model win, I suppose. Clearly I need to work with with Aster on a deeper level, i.e. at all, because I have not.

00:15:53:15 - 00:16:13:01
Unknown
But you were busy this week yet. Not dug into it yet, but I will, I will apparently it's. Yeah. Like I said, OpenAI says the world's best computer use model. Of course they're going to talk it out. I scored a 99.9% on Ark AGI three, which is a benchmark that was basically designed to test for AGI level reasoning.

00:16:13:01 - 00:16:46:20
Unknown
Okay. Also, what did it get a perfect score on an exploit bench, which, you know, is the first LLM to meet their critical cybersecurity threshold. So, you know, it's it's super powerful and and all the things. And we've already kind of been talking about it in previous weeks. Right. Because it's kind of at the core of, of a lot of the, the breakout stuff that OpenAI has been that's been revealed around OpenAI.

00:16:46:21 - 00:17:08:26
Unknown
Right, right. Like the, like the German story from last week. Is that also related to Astra or is that just another. No, that's that I think that was prior. And whether that was part of the problem here is we don't know whether any of these things broke out because we don't know what the original prompts were. Right? So we don't know what the restrictions were.

00:17:08:28 - 00:17:19:10
Unknown
We know it was given a task and tried to meet that task. That's what it's going to do. And to use words like break out or hack.

00:17:19:13 - 00:17:41:01
Unknown
Is problematic, I think. And the other thing I was told to do a thing and it did the thing. It found the the entryway, it found the door or whatever it according to Leo Laporte and the discussion offline with him, I think he said that the process in the original hugging face air quotes hack was something like two weeks.

00:17:41:03 - 00:18:10:13
Unknown
So the irresponsibility of not checking on it, not seeing what was happening, not knowing what was happening ongoing, not having a means to check on it is another issue here. It was given a task, it went off and it went this and that and did this and that and then and then the Cal Newport, I think it was his podcast, said that when we use words like swarms, it becomes kind of emotional, but it's really just one process, hiding off another process.

00:18:10:15 - 00:18:27:26
Unknown
And then that process comes back to the to the master process and says, okay, here's my here's what I'm delivering to you. So it's not as if it's these, you know, 1200 calculating agents going out there saying what problems can reveal. What have we got to do today? What sites are we going to hack today? Come on. The wicked Witch of the West.

00:18:27:26 - 00:18:53:16
Unknown
Monkeys flying around, right? Yeah, yeah, it was just just on DNS. Daily tech news show with Tom just right before this show, actually. And we were kind of talking about that and just talking about the difficulty, how difficult it is to not not just default to talking about these things in human terms, you know, like, like we talked about a couple of episodes ago, which I thought was a fantastic discussion.

00:18:53:16 - 00:19:16:05
Unknown
It was very eye opening to me as far as why you don't want to do that, why it's a bad idea to humanize these systems that way, not to mention inaccurate. But it's really hard, you know, to because what is related to this German wiki thing, you know, when you get into the details, the, the avenue it found.

00:19:16:07 - 00:19:39:04
Unknown
So I'm going to I'm going to horribly slaughter this. But but of my understanding is that these agents were given access to git. You know, the git command, which is essentially it's like a retrieval command, however. And so they shouldn't have had the keys to right, which essentially they used in order to like communicate to each other or even maybe even communicate is the wrong word.

00:19:39:04 - 00:20:15:22
Unknown
But to report the status of what each agent was doing to make these changes to this wiki and the git command shouldn't have given them the ability to do that. But what they ended up finding is that the wiki was hosted on a server, like a Python server or something from like 20 years ago, and it had some sort of capability built into like the address, the actual address that you, that you type out that if you put on this extra extension to the address that allows you to actually edit the site.

00:20:15:22 - 00:20:56:22
Unknown
And so it sits outside of the git commands capability of writing, it's baked into the site itself or into the code itself. And the agents were the ones that discovered that, that that, you know, did whatever they were given the task, they did whatever searches or queries or what to eventually come across that piece of information, and that ended up being their way in which to us looks like, oh, well, they hacked into the system, they figured out how to write, but it's it's not that they just did the right search to find the piece of data that we as humans might not have found or thought of or whatever.

00:20:56:22 - 00:21:16:05
Unknown
And so my point in talking with Tom is the term I used was crafting. He's like, but it's not crafty because that's a human thing. That's a human right like terminology being applied to it. And like I recognize that. But at the same time, like that's, that's the word that I have for man. I, you know, humans might never have thought of that, but yet they did.

00:21:16:05 - 00:21:43:01
Unknown
That feels crafty even though it's not. I don't know, it's weird. I think we discussed this I think last time on this. Yeah, I'm struggling with that. What I started, I seen calculate not think right. I think calculates this or that. Then it starts a process that's that's setting off an agent. Yeah it.

00:21:43:03 - 00:22:01:26
Unknown
I don't know what, what else we need to do but I don't it the worst thing to do is to ascribe motive to it. It has absolutely no motive nor a sense of right or wrong, nor a sense of crime. So it doesn't. It doesn't know what it's doing, something wrong. It doesn't have a sense of right or wrong.

00:22:01:26 - 00:22:26:27
Unknown
Now, it can violate its instructions. But is that is that the instructions fault for not being clear? Right. It's not like they're like, all right, let's violate some instructions today. Like they were given a command and that that violation may or may not have been, you know, have just organically come up in the midst of, of them. God, I don't even know how to talk about this stuff.

00:22:26:29 - 00:22:50:01
Unknown
It's really hard to talk about it without human terms. So in terms of the skill, I think people understand that it's just a prediction machine for the next likely word. Boom, boom. Yeah, right. Yeah. In terms of the input, it's the same issue. It takes it takes whatever command you give it and it tokenize it. It turns it into just tokens with no sense of meaning.

00:22:50:03 - 00:23:19:03
Unknown
Yes. So I think that we keep ascribing meaning to these acts, that our instructions are clear because they can describe meaning. It can't even do that. Yeah. Now it's still amazing what it does with those logical limitations. Phenomenal. But it cannot think. It does not know things. It does not know the meanings of the words. It knows the knows it.

00:23:19:05 - 00:23:52:26
Unknown
It calculates the associations of words and the proximity of words. It maps those things. It uses those associations to infer meaning for the purposes of functionality. We need a drier language like that. Yeah, I was just kind of sitting there in or listening to you so easily spell it out in those terms. That is that that takes an educator like it.

00:23:52:27 - 00:24:16:08
Unknown
It will take an educated and very deliberate kind of rethinking, but it's not going to get you talk about these things. No. It's so much easier for people to anthropomorphize. It's much more, well, especially in it, especially because the vast majority of people that any of their interactions with these things are as if it is a, you know, they call it a chat bot, right?

00:24:16:09 - 00:24:37:08
Unknown
Like it's it's something to chat with, to converse with, to talk with. And it's like a representation of a human on the other side of the glass, which is an abstraction that we're very used to in talking with actual humans. So what is the difference, you know, to convince people that there is a difference? There is a is a tall order, I think.

00:24:37:10 - 00:24:44:11
Unknown
Yeah, yeah. Interesting. And then what about the.

00:24:44:14 - 00:25:09:26
Unknown
So much of of what we understand about why these models decide to do what they are doing is because of kind of like the paper trail that they leave behind that like reasoning paper trail. And I feel like I've started to see stories about future models not leaving that paper trail. Well, there was some there was some talk that Astro was not going to, though, that got misinterpreted, I'm told instead.

00:25:09:28 - 00:25:34:05
Unknown
And so this is interesting too. And I've said this for a very long time. When we came to what we think of as reasoning models, we came to think of it as that, because it would slice up a task and tell us, I'm doing this and this and this and this and this, right. So I give Gemini a task about finding me statistics about something, or I had it because it's so hard to get book reviews these days.

00:25:34:05 - 00:25:47:21
Unknown
I had a write a review of my book just for the fun of it, to see what it did right. And the you see that stuff flashing by, I'm doing this, I'm doing that, I'm doing this, I'm doing that right. Using anthropomorphizing.

00:25:47:23 - 00:26:16:02
Unknown
Again, I go back to a podcast he normally writes about using your full mind and other things, but but he also writes about the technology or talks about it. And the problem is, and I've said this for a long time, I believe that those indications of activity or post facto are kind of made up to satisfy us, that they're not necessarily indicative of actual tasks.

00:26:16:04 - 00:26:55:03
Unknown
And so how valuable is that script in terms of understanding how the how the program, how the software went through its process? I don't know, which is to say that something more is happening underneath. Yes, underneath that is not being detailed like like what you see in that, that reasoning thread or chain or whatever they call it isn't the extent isn't the entirety of the conversation or the way is that there's, there's sub, there's sub layers of it and or that stuff's all made up.

00:26:55:04 - 00:27:13:25
Unknown
And in fact, the actual sequence of I tried this, I tried this, I tried this, hope that worked, I tried this, I tried this, I tried this, hope that worked that that's not legible to us. And and so that's the concern people had with Astros. If they're cutting down some of that. Well what are they cutting out. Are they clean out that fake layer.

00:27:13:26 - 00:27:32:11
Unknown
Are they not a real layer. How legible is that real layer. Is that what it's really giving us? In truth we don't know. And going back to the hugging face incident again, we don't even know the prompts. So we don't know what instructions it was given and what it was forbidden to do. And did it jump over that?

00:27:32:11 - 00:28:00:10
Unknown
And how clear were those instructions? And on and on and on and on and on. It's a you know, we used to use the Wild West term a lot in the early days of the internet, but these are kind of unbroken ponies. They're just running around doing stuff, and they haven't been trained in our sense of knowing that they're reliable in what they do and that we can understand what they're doing.

00:28:00:13 - 00:28:04:28
Unknown
And that's difficult. Yeah.

00:28:05:01 - 00:28:32:02
Unknown
So I don't want to go in this in detail. I added it into the rundown. But the chief scientist at OpenAI wrote a long post saying that calling it an alien mind. Jacob, if I'm pronouncing that correctly, chief scientist, OpenAI. And so he's just going on about an intellect we don't fully understand. So a to call it an intellect is a leap.

00:28:32:04 - 00:29:23:14
Unknown
To call it intelligence is a leap. It is a software program that, with sufficient compute, can do amazing damn things. Right? But it's still software. And, you know, it has. I'm just reading the headlines here. Teaching machines to love. Gag me. Scalable defense. You know what's next. So I think that they're going overboard in all kinds of ways here that we've got to break away from these these AI boys and put the I really wish that we had a government that would fund basic research on this stuff at universities, where they have higher standards of evidence, where we can better understand what's going on behind, so we can get some some visibility on these

00:29:23:14 - 00:29:46:16
Unknown
processes because we don't have it now. No, we really don't. We are left what want waiting for them to let us in on how they do these things. So I don't know whether you want to play this or not. I just put it in. Just as we got on the trailer for the OpenAI movie about Sam Altman's firing just went up.

00:29:46:19 - 00:30:08:15
Unknown
Oh boy, artificial boy. I wonder if I'm allowed to play it. I wonder if I play it, if I'm going to get, well, maybe don't. So. So you can you can find it in the. Or you can play without sound. Let's just. Yeah, I could play without sound I think. Yeah, yeah. But it's in Silicon Valley. A very expensive car.

00:30:08:16 - 00:30:35:16
Unknown
Okay. It's a Batman car. Is this a Batman trailer or is this kind of. Looks like it was. Oh, I mean, from Sam Altman walking. Yeah. Definitely has the Sam Altman look to him. Okay. All right. Sure. Yes. Going down into his bat cave. Maybe he is actually Batman. I'm not really quite sure. Rifle. Okay, so how much of this is guns?

00:30:35:16 - 00:30:55:01
Unknown
I didn't know there were lots of guns. I know machine guns, and. Wow, this is this is going to be real interesting. I have a feeling. Parties, women. Okay. Yeah. All right, so it's out. All the standard stuff. It'll be out Christmas Day. What a great Christmas gift. Go watch them. Want to get fired. I doesn't strike me as a Christmas movie.

00:30:55:03 - 00:31:20:10
Unknown
No, no, not so much. Maybe. Maybe to the AI boys, though. Yeah, maybe. And then. Oh, real quick, since we're talking trailers, did you see the trailer for the Theranos documentary thing? It's amazing that one. That one is. I don't even know what to expect clicking on that, I just saw the the trailer like a thumbnail or whatever.

00:31:20:13 - 00:31:54:28
Unknown
Right. Anyone out there, you just gotta go see it. There's a there's an upcoming documentary coming for, which is like the last days before. What's her name? Holmes. Elizabeth Holmes. Thank you. Her 11 year sentence in prison. And it's like the days leading up to her prison sentence. David gave access to a filmmaker and a filmmaker who makes unusual stuff, and it was so, I don't want to say shocking, but so filled with surprises that when they screened at Telluride, you all had to leave your phones outside.

00:31:54:29 - 00:32:22:07
Unknown
They scolded everybody. This is a big secret. You can't spoil anything, blah blah blah. But people were like, whoa, because she's such a odd character with these huge eyes that kind of never blink. Yeah, I know, she kind of looks like an avatar, right? Yeah, like she just kind of has that quality to her. Like a what is the what is the the term where something looks not quite entirely uncanny Valley uncanny.

00:32:22:08 - 00:32:41:26
Unknown
She's got an uncanny valley quality to her. She's always kind of had that a little bit. And then this trailer, like it's pretty much mostly just her like and him looking at each other and, and it's just very quiet and it's definitely worth. This is the director, right? The guy is the director. Is that right? Oh. Where is it?

00:32:41:26 - 00:33:06:16
Unknown
I'm not seeing the name. Yeah, I'm not sure anyone worth. So it comes out about the same time as The Social Reckoning, which is the sequel to The Social Network, a movie I hated because I think it made up stuff. So I think, I think this is going to steal the thunder. Yeah, yeah, it could, could indeed. I had no idea this was even a thing.

00:33:06:16 - 00:33:25:26
Unknown
And then when I saw this, I was like, I have to watch that. I watch a documentary. There's already been a mini series about her. There's already been a show about her. I think people thought, oh, what's the big deal? But now this is this has made huge buzz and it's a documentary. People don't watch documentaries. Yeah, it's not a big a big world.

00:33:25:26 - 00:33:46:15
Unknown
So. So it'll be. Yeah I can't wait. That's that's worth the ticket. And I'm buying the popcorn. Indeed. Get your popcorn ready. It's going to be a show. Yep. Speaking of a show, we do a show every single week. Sometimes we do it a day early. Sometimes I do it five days a week. If you're a patron, because I do a shortened version of it.

00:33:46:17 - 00:34:00:29
Unknown
And which, by the way, if you subscribed and you didn't get one yesterday, it was Labor Day in the US. And so let me give, give a break, give a break. I could, I could have I could have given you the warning. But I was so busy being on Labor Day that I was just like, you know what?

00:34:00:29 - 00:34:21:15
Unknown
They'll figure it out. Yeah. So I took the day off. Anyways, Patreon.com show is where you can go to support everything that we do. We've got a number of things that we do here and it's all supportable there. And we do have some amazing patrons like William Wilkinson and Marie Texier. So thank you for supporting us. Thank you.

00:34:21:16 - 00:34:43:26
Unknown
Week after week. See, I told you I was traveling. This was from my hotel room in Arizona. I literally I had ten minutes. I was like, we had ten minutes in between things. And I was like, I gotta get this out. So I threw it together while I was like, traveling around. While we were traveling around, I was like, chatting with Claude, like, okay, these are the stories and blah, blah, blah.

00:34:43:26 - 00:34:59:20
Unknown
And then I got home, fired it up. So this was a college visit. Was it also a tournament or something or a meet? No, it wasn't a tournament. It was a college visit. It was college. It was. We think you're a good swimmer. We want to fly you out and show you our schools so that you'll consider coming here.

00:34:59:22 - 00:35:18:19
Unknown
Did you have to swim? She. She couldn't swim for them. The rules actually specify that you can't do that. But she did swim with the team. They did. It's cool. Yeah. The team was, you know, like, pulled her in and, like, her and a few others. She wasn't the only one there. There were a few recruiters. That's very cool.

00:35:18:21 - 00:35:37:09
Unknown
Yeah, it's a cool experience, I gotta say. I'm really enjoying it. It's cool to see her in such a, I don't know, in a position where multiple teams are like, we want you, we want you here. Come on out. We'll take care of you. We'll show you what it's all about. Cool experience. But anyways, I even do the the daily podcast from there.

00:35:37:13 - 00:35:54:22
Unknown
From a hotel room with no time. That's right, that's right. Were you wearing pants? You don't need to know the answer to that question. Look, it's Arizona, all right? It was like 100 degrees. You do what you got to do, Jeff.

00:35:54:25 - 00:36:10:06
Unknown
All right. And with that, let's take a break. When we come back, we've got a few, like, newsy stories, but then a bunch of just very interesting stories in the world of AI. So some pretty fun stuff coming up in a moment.

00:36:10:08 - 00:36:37:24
Unknown
All right. I wasn't sure whether to include the Google ad tech antitrust ruling on Friday. This is basically the Department of Justice. I mean, essentially saying like Google, you don't have to break up your ad tech business. You got to make some changes, but your ad tech business isn't going to be totally and completely disrupted. To which, you know, I'm sure people have all variety of opinions on this.

00:36:37:28 - 00:36:56:01
Unknown
Is this an AI story? I feel like it kind of is because because Google is such an AI company. Yeah. And so it's how they're going to how they support their work or the advertising. It's just like, you know, Amazon's and all kinds of legal stuff now. Well, Amazon hosts a huge amount of AI that we have. So it's at that level of technology.

00:36:56:01 - 00:37:23:02
Unknown
It's intertwined. Yeah. Yeah. You know and OpenAI is an ad company. Now, that was part of the rationale was that there's lots of new competition to Google in advertising. Now that's Amazon, that's also open AI. I still don't believe it. OpenAI has $1 billion run rate of advertising. But we'll put that aside. And yeah, it all it all matters in the big tech hegemony.

00:37:23:04 - 00:37:45:06
Unknown
Yeah I mean yeah, right. ChatGPT ads reaches $1 billion in annual revenue run rate. That was last year. I don't believe it. You don't believe it? I don't believe it. You don't buy it? No, because. Because it takes a lot of time to build up advertising. And you need sales and you need support and you need data and you need other things.

00:37:45:06 - 00:38:03:01
Unknown
I don't see advertisers saying, oh, is really working for me out there, and a bunch of stories in the media world. So I don't know how they're coming up with that number. There's some Gary Marcus has been going nuts with listing and not wrongly. I don't mean to say that he's nutty, but I'm just saying that he gets excited.

00:38:03:03 - 00:38:27:28
Unknown
Listing OpenAI's lies and prevarication. And I don't know if he list this one or not, but this one strikes me as suspect. I mean, that would be a pretty quick, quick build up. And but it is important that Google does not have to break up the ads. It has to provide some more, as you say, transparency to the publishing clients here.

00:38:28:01 - 00:38:54:19
Unknown
This is the second attempt to break up Google and the second attempt that's failed. So I think that matters for Google's ability to remain a big company in this. And the other thing, Jason, yeah, there's a sigh of relief. I just sent in a draft of a piece for Project Syndicate about about AI and privacy, but in that I write about data as a whole, and I have to include discussion of the attitude toward AI companies.

00:38:54:19 - 00:39:30:03
Unknown
And I think that there's a special spot in people's hell these days for the two air quotes, frontier companies, OpenAI and anthropic. And I'll make it three throw in space. But especially the first two are kind of pure AI companies. Google and Amazon are not pure play AI companies, but they're becoming majority AI companies. But they have enough diversification to do other things and enough diversification that people don't necessarily hate.

00:39:30:06 - 00:39:53:28
Unknown
When they think about hating AI, they don't necessarily think of Google. They don't immediately. I mean, there's plenty of there's plenty of AI. Hate to go around for Google. Yeah, yeah. But I don't think they're leading the pack there. No, I think you're right. I think yeah, yeah. It's easier to throw that on the companies that that is clearly the, the only thing they have going or the basis around which everything is, is built upon.

00:39:54:00 - 00:40:19:07
Unknown
And Google's been around long enough to have already built a basis of a lot of things. Now it's squishing all the AI into what they already have. Yeah, in all sorts of different ways. But yeah. Yeah. And so much of Google's ad business, you know, isn't just that like ad marketplace anymore as well. Right. Like it is about smart, you know, AI driven bidding.

00:40:19:07 - 00:40:58:07
Unknown
It's about AI generated ad creative tools and all that kind of stuff. So their ad tech does extend beyond just like the legacy ad sales business and into the AI tools that influence it so well, they're breathing a big sigh of relief, that's for sure. Yep. Google's that's that was a champagne pop victory I'm sure for Google. Atlas this is this is a cool example, I feel of, you know, once again of AI in a direction that could be really beneficial and useful and helpful for a lot of people.

00:40:58:08 - 00:41:33:25
Unknown
Potentially down the line is Google DeepMind releasing Alpha Genome Atlas? It's an AI powered database that's built on its alpha genome model, predicts how all all 9 billion possible single letter changes in human DNA could affect biology. And so, you know, it takes a look at all the different ways that those variants can, can take place. And then out of those variants, out of those, those ways in which, you know, it could go, I don't know, with different combinations in different directions, that sort of thing.

00:41:33:26 - 00:41:56:19
Unknown
They can monitor for those mutations and then identify the things that would be, you know, the discoveries that would be most beneficial essentially. So the human genome says Google is made up of 3 billion base pairs of DNA. Scientists understand the 2% of the human genome genome that codes for proteins relatively well, but only in limited knowledge of the remaining 98%.

00:41:56:21 - 00:42:14:16
Unknown
So the fact that Google's Alpha Genome Atlas is a database that predicts the effects of every possible single nucleotide variant in the human genome, and they're making this open and they're putting this out of the world. I think I think you're right. This is this is where AI and tech companies do well.

00:42:14:19 - 00:42:42:26
Unknown
Yeah, I won't understand it. That's pretty remarkable. Yeah. I'm not the person to work with it I'm sure. Yeah. But it's but the fact that this sort of resource is being open, like what does that lead to. That's that's pretty remarkable. The related news, I just add that out of this, below that a connectome milestone, Google has mapped the complete male fruit fly brain.

00:42:42:28 - 00:42:46:22
Unknown
Wow.

00:42:46:25 - 00:43:13:12
Unknown
So lots of work going on here. Wow. What do you what what can you do with a with a flew a fruit flies brain? Once you know exactly how it works, work up to the next ones. Yeah. Yeah, exactly. It's a good starting point. Oh, that's that's very pretty. It compliments the female fruit fly brain map and recently released complete female fruit fly brain and nerve cord map.

00:43:13:14 - 00:43:26:15
Unknown
Wow. So next just next. What are they doing next, folks? Fish brains. Fish brains. You know, still still not very smart.

00:43:26:18 - 00:43:48:02
Unknown
Animals. Beings. What do we want to call it? But still, you got to start somewhere and imagine this complexity. Like, I don't actually know what I'm looking at, but I'm assuming this is a visual representation of the map itself, which is in itself just incredibly dizzying with the amount of detail and connections and all that. And this is just a fruit fly.

00:43:48:09 - 00:44:11:03
Unknown
Imagine how complex that gets when you scale up. That's crazy. That is super cool. I missed that one. Thanks for putting that in there. That actually, you know, that ties in with I'm going to move around the order of something here because you just threw that on to us. And we've got talking to animals using AI to talk to animals.

00:44:11:03 - 00:44:57:01
Unknown
This is this is really interesting. Researchers using machine learning to basically decode how animals communicate. They've compiled 150,000 recordings of Carry On Crows. And they're just and I guess they're just using the artificial intelligence models to analyze what we know about their communication. And potentially we get to a point to where we can actually communicate with animals or at least understand what they're saying, possibly communicate with them, which could be beneficial, could be useful, could also be abused in a lot of a lot of kind of negative ways.

00:44:57:03 - 00:45:22:13
Unknown
Yeah. Very fascinating. Yeah, I think it's fine. It's to me, it's a little weird that they lead with this notion of the ethical minefield, but fine, caution always warranted. That's good to have. It'd be pretty amazing. But I think that there'll be really difficult conceptual, conceptual questions here because they don't you know, dogs have those. Don't know if you've tried this with yours, with that pad with the words on it and it speaks.

00:45:22:18 - 00:45:49:24
Unknown
Yeah. Never tried that. But I've always had questions about it. Like really, really real or not. Right. But in that case the dog is learning a human connection. Yeah. Or I've seen where deaf dogs respond to ASL signs like they respond to words. They are adapting to us to turn it around. How does a dog think? We have no idea at all.

00:45:49:24 - 00:46:19:26
Unknown
And if we try to put it in our terms, we've spoiled the the worldview, right? You've got to be able to think how they think. And I don't know that we know how to do that. So dolphins we know communicate. Birds communicate. Dogs communicate in ways that are besides just oral. Does your dog do the play stance thing when they put their front paws down on a stick up, right?

00:46:20:02 - 00:46:48:10
Unknown
Yes of course. Yeah. How that became a universal for I'm not going to kill you. Let's have fun. Right? Yeah, yeah. So there's lots of communication cues that go on. Oh, for sure that we may not know. So AI I think is going to need video views of this. Not just audio, not just brain. Not just. Yeah. Well and so the article like you said kind of goes into some of the ethical kind of considerations here.

00:46:48:12 - 00:47:23:09
Unknown
Right. And you know, it points out, you know, when you're playing back recording of animal calls, for example, that can cause stress. It can disrupt, you know, like their hierarchies. It basically kind of interrupts their natural living environment and causes them to change behavior or whatever. And so the ethical side of this is, you know, when we can talk to animals or when we can communicate to animals, there will inevitably be people that use that to do, you know, not great.

00:47:23:12 - 00:47:48:28
Unknown
There's event killer dolphins with with nuclear arms on their back. Right? Sure. Yeah, right. Spy dolphins, that kind of stuff. There's great. My favorite stories out there, the ones about crows being very smart. But yeah, animals adapt to us in all kinds of ways. Bears come to do right. You know, crows figure out how to bribe us or how to attack us.

00:47:49:01 - 00:48:11:00
Unknown
Yeah, we live in an ecosystem with these, with these other living beings. Well, we all. Yeah, like like like you just said a few minutes ago, we already kind of communicate in a certain way with animals as well, right? Like, like some people are just really good at making animal noises with their mouth. And when they do that, those animals come near.

00:48:11:00 - 00:48:34:22
Unknown
Or hunters, right. When we go hunting, people might get that, that bird call, you know, device at the, at the store and use that so that they call the right bird pulled up. They've, they have they have hacked the system essentially to trick the animal into coming over. And so it's it's interesting, when I was reading this, I was thinking of that and I was like, not that I think that it's right.

00:48:34:22 - 00:48:57:13
Unknown
But like, where do you draw the line between one and the other? Because we're already doing that over here. And now suddenly it's not okay to do it over here, even though the technology is advanced more, you know, I don't know. It's interesting, just this idea that we could someday talk to animals to be so cool. That would be a Doctor Dolittle.

00:48:57:15 - 00:49:19:27
Unknown
How how would we know without a shadow of a doubt, that what we thought they were saying was actually what they were saying? You know, that's the. Yeah. Because once again, they will try to adapt to what we're doing. Yeah. You have to be in a very virgin situation.

00:49:20:00 - 00:49:52:02
Unknown
Fascinating. Back to Google. And in news that you I just I read this and I was like, well of course, like of course three hikers from Roseville, California decided to climb Mount Shasta. They use Google Gemini to plan the trip. Gemini told them what could possibly go wrong. What could possibly go wrong? I know they told them to bring food and water, but actually less than they might need for the climb.

00:49:52:04 - 00:50:19:06
Unknown
They set up camp at 8400ft on Saturday. Sunday morning. They left at 3 a.m. with day packs, you know, built around what Gemini told them essentially summited at 7 p.m. and the the actual recommended time was to Shasta or on Shasta was noon. So they were seven hours beyond that essentially. And that meant that they had to do the go all the way down in the dark.

00:50:19:07 - 00:50:45:19
Unknown
They're going down. They don't have enough food. They don't have enough supplies. One guy injures his knee, they wander off the route. And ultimately dispatch was called in and USFS climbing rangers found them the next morning, treated the injury and escorted them out. So the moral of the story is, don't use Gemini for something as critical and important as something like call Ranger Rick, call the Ranger instead.

00:50:45:20 - 00:51:11:18
Unknown
Yes. Yeah, not a good idea. It's one matter if you're asking. It's cool. Places to see in Rome. Okay, that's all right. That's all right. But not where your life is. That. No, don't don't do that. No. So horrible, horrible idea. And not to mention, I mean, are you really going to know when it gives you advice that is good, good advice versus bad advice.

00:51:11:24 - 00:51:38:00
Unknown
If it tells you to jump off a mountain, you know, jump off a cliff or whatever, you're going to read that and you're going to obviously you're going to know, like, oh, that thing's hallucinating, or that thing is telling me something that is, it shouldn't be the case. I should not do that. But if it's within the realm of, you know, just enough makes sense, just enough, and you don't know enough to question it, like, yeah, it's just a bad idea.

00:51:38:03 - 00:52:08:21
Unknown
Yeah. Don't do that. Yes. Of course, everybody you know, in response to that story is I would never do such a thing, but, Yeah, don't do that. I also thought this was interesting because we've talked about, you know, kind of tokens, the tokenizing of the world and how tokens are becoming more and more. Well, I noticed it at my at the workshop that I did a couple of weeks ago that people were mentioning tokens a lot, and it was like, okay, there's an awareness, there's a growing awareness of what a token is.

00:52:08:22 - 00:52:40:18
Unknown
And apparently in China, tokens are kind of taking a life of their own, taking on a life of their own in culture, in the sense that a lot of a lot of different businesses and companies are using tokens, that is AI tokens as kind of incentive or as payment for certain things. So if you sign on at a Bank of China, basically they give you out Kimi Kimi AI tokens as credit card rewards.

00:52:40:22 - 00:53:07:04
Unknown
So now we're we're starting to get these tokens almost kind of living up to their name there. You know, it kind of feels like money, which it kind of is because when you use a token, it does cost you money, right? Yeah. It is an economy. So there's credit cards here. There's one for the China merchants Bank was launched, credit card for ad developers, up to 1.8 billion tokens for use in Chinese AI startup minimax models.

00:53:07:12 - 00:53:31:28
Unknown
Another bank had won subsidies for 3 billion tokens. Then there's mobile data because phone companies get on this. China Telecom offers consumers plans starting at 9.91 $1.40 a month for 10 million tokens. AI tokens as a freebie, as a premium, I worked at Time Inc in the day when you'd order Sports Illustrated and you got the sneaker phone.

00:53:32:00 - 00:53:49:15
Unknown
We ever talked about that? It was a sneaker. That was a phone. You held the bottom of the sneaker up to your face, and it had a cord coming out of it, which of course. Okay. The case, like I've seen a sneaker phone. So the speakerphone was famous. They used to order time magazine. You get a get a clock.

00:53:49:18 - 00:54:13:01
Unknown
These are premiums. So premiums for restaurants, cafes are luring in customers. We had I think we had a story the week about the AGI bar, where customers by a drink and connect to the Wi-Fi get a limited access to deep seek V4 flash there, but they're not making any money from it. AI tokens in e-commerce and in bank lending.

00:54:13:03 - 00:54:18:12
Unknown
So I think we're going to find.

00:54:18:14 - 00:54:40:12
Unknown
It's weird to me that it's talking. What if what if somebody said, we're going to give you an hour free on Claude if you buy this or that? Okay. Yeah. It doesn't really work without context. Yeah, totally. Are you already on Claude? Are you already doing things? There is an hour of free useful, I don't know. Yeah. Is is that the model that you choose to use?

00:54:40:13 - 00:55:03:01
Unknown
Exactly. But, yeah, this is this is the way, trends build. Maybe we need a universal token. We need a universal token. Yeah. And it's, you know, we'll call it Bitcoin. No, we won't call it that.

00:55:03:03 - 00:55:23:25
Unknown
A universal token that could be applied to any of of the models that exist. Yeah. Just yeah, it just kind of reminds me of back in the day when, when coffee shops, you know, started to become WiFi hubs. Yeah. And that was the lure to get you in there. Come on into our shop. We've got free Wi-Fi you can go to now.

00:55:23:25 - 00:55:43:26
Unknown
We've got we've got knowledge in the form of tokens. You don't have to use your own content of the day. So kids, kids go to McDonald's parking lots. Poor kids do because they get Wi-Fi there. Well, they also get AI there. And it goes to the story we had last week about South Korea giving everyone free. Totally. Yes.

00:55:43:26 - 00:56:08:12
Unknown
Yeah it does. Interesting. This next one is disturbing for me because hilarious it is a thing right now is that people are noticing and I've certainly noticed it. I don't know if you've noticed it out, out and about in the real world. Kind of an uptick in using AI for restaurant menus and restaurant marketing. What's wrong again?

00:56:08:12 - 00:56:34:00
Unknown
Jason? Oh my goodness, this is just. Yeah, all the food looks the same and it's all glossy and perfectly round and perfectly drippy and doesn't look and not related at all to what you're going to end up eating. Thank goodness. Like, what is that? It's protein. Oh my goodness. It's like a it's like a worm or something or I'm not entirely sure.

00:56:34:00 - 00:57:01:16
Unknown
But yeah, there's some really disturbing stuff. I had come across this photo a couple of weeks ago of a burrito. And sorry for audio listeners, but, you know, there's links in the show notes at AI inside. It's like a weird sponge slash virus. It this image disturbs me, and I think I finally understand the people who are who are afraid of of holes.

00:57:01:19 - 00:57:25:19
Unknown
What is that? That it starts with a P. It is a phobia. It is a phobia. Phobia. Right. Which is the fear of like sponges, like colonies of holes and stuff. And I look at this and it just makes me think of like alien food, like, imagine visiting another planet and they give you this. Do you eat it?

00:57:25:21 - 00:57:50:20
Unknown
No, no, I don't I don't put that anywhere near my mouth. That is just scary. And then and then I also saw these these weird looking shrimp things, which I'm not entirely sure what I'm looking at. It's meant to be shrimp, but oh my goodness. Just don't don't don't use the models for your food. No food ads us just disgusting.

00:57:50:20 - 00:58:11:01
Unknown
It's very unappetizing. And I mean, you know. And so then of course, along with this, you know, the Guardian had the article that that kind of made me put it in here. I've been thinking about this for a while, but part of the reason why AI food looks the same is because it's kind of like the food itself, the images of the AI food, it's it's trained on itself.

00:58:11:01 - 00:58:51:12
Unknown
And over time it gets more and more just kind of I don't know if it's perfected or glossy or what it is, but it transformed what what it believes, what the model interprets as an actual representation of food is kind of tainted by the fact that there's all sorts of AI generated food out there. And it you know, it's like running if you if you use ChatGPT image and you put a picture of yourself into it, and then you keep feeding that picture into it to have it reproduce that picture, the the the one picture to the next might not look very different, but but go forward like 20 or 30 photos.

00:58:51:12 - 00:59:12:18
Unknown
If you've ever seen people do this, it's really interesting and it changes dramatically between the beginning and the end destination, because it's kind of like continually basing it's, you know, it's work on the last. It's one of the books that be publishing for intelligence, AI and humanity is Matthew Kirschenbaum lips when he talks about this? Yeah, that's exactly what we're talking about.

00:59:12:19 - 00:59:34:16
Unknown
It feeds on its end trail and it turns into this gray goo. The same with images. It feeds on it. So all burritos look creepy. And how I don't want to eat a lowest common denominator. No, I will never eat that burrito. No, and I will never eat those shrimp. I don't even know what that is. But whatever it is, I will never eat it.

00:59:34:18 - 00:59:53:28
Unknown
They're they're like very big sea maggots. I mean, that's kind of what it looks like. It looks like, like really gross, like ringworm. There's a lot of them because nobody wants to eat them. They got plenty. Oh, but you'd you'd be so full if you're one. And only requirement is I want to be full at the end of this meal.

00:59:54:01 - 01:00:17:02
Unknown
Get this. But if it's anything else, do not get it. One little slice of lime tries to add flavor to the maggots, sea maggots, and a couple of sprigs of basil in there or something. What's the quantity? Because this is America. Are you have all these all these shows with this ridiculous quantities of food, these huge sandwiches, which I hate.

01:00:17:03 - 01:00:35:21
Unknown
I think it's just so American. But that's kind of infected this because the quantity each of these cases, it's a lot of food. It's massive. Yeah. Yeah I mean if if this is like chicken wings you're eating like yeah I guess with chicken wings, if you're ordering wings there's a fat wings. Those are. Yeah like turkey that weighs.

01:00:35:22 - 01:01:00:18
Unknown
And they just does not look appealing. See it's lunchtime for me and I don't want to eat right. I don't want to eat. Break. Yes. Let's take a break. I'm not going to go eat some lunch during the break, I'm gonna think about all the horrible choices I made in my life. And when we come back, we'll talk about a few stories in the speed round to get you out of here.

01:01:00:21 - 01:01:26:15
Unknown
All right. The smart glasses inning continues with meta as the company is permanently bricking cameras on Ray-Ban smart glasses when it detects that the LED privacy light has been tampered with. It is, I believe, less than 0.1% of all glasses sold. But, you know, you've got a lot of devices being sold out there. That's a that's thousands of devices, I suppose.

01:01:26:15 - 01:01:43:24
Unknown
But people don't don't, you know, don't hack your glasses to turn off the LEDs so that you can do stupid stuff. Don't. And if you do meta saying they're going to break your device. So keep that in mind. I saw a story today. I didn't put in the rundown of police worried that people are going to use these glasses to record them.

01:01:43:28 - 01:02:15:03
Unknown
Well, yeah, yeah, yeah, you'd have a good idea that they were because they'd be wearing them to. So yeah, I mean, people are already recording police doing bad things. They wouldn't know it. That'd be sneakier. Smart glasses. Yeah, yeah. Okay. I guess slightly sneakier. They're pretty easy to to figure out. It all gets interesting. So I saw another story that said that if.

01:02:15:05 - 01:02:42:16
Unknown
We had an age cutoff and time cut off on social media, we would never have seen the George Floyd video because it was a teenager who did that. Right. So so this question of people being able to record things that happened in public places, there's a lot of nuance to that discussion. Absolutely. There is. Yeah. Yeah, I totally agree.

01:02:42:18 - 01:03:18:03
Unknown
Microsoft announcing Project Zenith. This is a specialized Windows 11 setup for developers and it's pre-installed with VS code, GitHub, Copilot, Power toys, Python, node 24, but things. Basically it requires 64 gigs of Ram so that you can run AI locally, among other things. So there you go. Coming from Microsoft, its own really expensive I'm imagining. I imagine this is probably a pretty expensive device, but yeah, we're gonna we're gonna see so many of these, aren't we.

01:03:18:05 - 01:03:39:24
Unknown
Like but but it's still the the level of expertise needed to run models is still high. I think that's going to change pretty rapidly. I think somebody is going to turn into something retail. But most of these companies you're training on it at at the server level, the cloud level. And I think we're going to find something that comes out that feels just like an app.

01:03:39:27 - 01:04:07:23
Unknown
Yeah, it really isn't that far away. It's just you have to go in no terminal to install things and stuff. I don't think it's going to be far away at all. Yeah. Well, and I mean everyday person, you know, my wife buying a new laptop, she's not going to she's not just going to happen into a laptop that just happens to have as much Ram and internals that would be required to run a really capable local AI model.

01:04:07:24 - 01:04:25:23
Unknown
You know what I mean? It's going to I mean, that may happen someday. That's not going to happen overnight. So this is still firmly in the you got to really want it in order to to do it. You've also got to really be able to afford it. So these things are so for basically it's 5 to $10,000 to run these things.

01:04:25:24 - 01:04:48:27
Unknown
Right. And expertise. But you know, like I was on the airplane of wanting to do some work and wanting to use AI models, and I was like, damn, I wish I had a local model installed on my laptop. Yeah, that would be really handy. But anyways, I'm not going to spend $10,000 on a computer though. Definitely not right now.

01:04:48:28 - 01:05:15:10
Unknown
Nope. Trump DOJ filed a statement of interest backing OpenAI's fair use defense in the New York Times copyright case. This is interesting to me because I'm not sure how often in this kind of private litigation, the Department of Justice enters into this extent. At this stage, they do friend of court and security briefs and such. That's that happens.

01:05:15:10 - 01:05:42:26
Unknown
Anybody can. But this strikes me as very interesting. Is the DOJ sticking their thumb on a scale? And does the court need to listen to DOJ? No, they're not on the instruction of DOJ. Would a jury no. So I'm not sure why they did this. Maybe because they believed the court will listen to them. Maybe. Depends on the judge.

01:05:42:28 - 01:06:11:21
Unknown
Depends on the judge. Right. And yeah, they they don't want to be seen as, as weak when it comes to AI. And so let's. Yeah I don't know. That's a good question. But there you go. Google had mentioned at IO a few months back that it was bringing AI voice features to Gmail to docs to keep and other apps within workspace, and that is now happening.

01:06:11:23 - 01:06:39:20
Unknown
So you can basically ask questions with your voice of all of these things instead of having to, I don't know, I forgot to try this because all I need to do is to say, hey, you know who and tell it to do something in my mail to see whether or not it will do it. Yeah. So I've got to try that, I don't know, I found, you know, when it comes to like, context, personal context inside of something like Gmail, I feel like using clod connected to my Gmail.

01:06:39:20 - 01:07:01:15
Unknown
I can do more inside of my Gmail than I can using Google's own AI tools embedded into Gmail, and I don't understand why that is. Like it's far more capable, or I get far better results when I use it as a connected thing through something else, as opposed to Google embedded tool, and I don't understand why that is.

01:07:01:16 - 01:07:21:11
Unknown
That's interesting. You know, I wish I think about this before. The thing I've found is that Gmail search sucks. Yeah, yeah, sucks even with the AI integrate. Well, so I went to the other day and I said, find me. I ran this event and I can't remember what it was, find this event that I did with so-and-so and so on and so and it found a recent discussion about doing another event.

01:07:21:12 - 01:07:33:21
Unknown
I said, no, no, no, no, go back. And it did, and it found it, whereas I would have had to search forever. Yeah. Just.

01:07:33:23 - 01:07:37:08
Unknown
Okay. Gemini.

01:07:37:10 - 01:07:41:03
Unknown
Okay. Google.

01:07:41:05 - 01:07:47:28
Unknown
Hey. Yeah. Hey, Google. Hey, Gemini.

01:07:48:00 - 01:08:13:15
Unknown
No. What am I, chopped liver? Yes, you are chopped liver. Oh well, let me show you an AI generated image that represents the chopped liver. Yeah, I don't like real chopped liver. Yeah, neither do I. Yeah, it could be a lot better. But anyways, I don't know. Maybe this will make it a little better, but I'm not entirely sold that it will.

01:08:13:21 - 01:08:33:28
Unknown
Yeah. All right. We've reached the end of this episode of AI inside. Everybody should go over to Jeff Jarvis. Com and find Hot Type. Buy your copy if you haven't already, or if you're waiting for an autographed copy, you know, just set yourself in. Yeah. Wait until Friday or set yourself an agent to go try to set up a structure so that if it's not in, you can at least leave your name.

01:08:33:28 - 01:08:53:15
Unknown
But they don't have that yet. They're very nice people. It's used. And a new bookstore in Montclair. They're great people. Hi, Ryan. Hi, Luke. And I've been selling I've sold out the books they have so far without trying to promote it. So I can't wait till they get enough books in and I can. Really? Yeah. Yeah, well keep trying.

01:08:53:15 - 01:09:15:08
Unknown
Everybody should go to Jeff Jarvis com, and you'll find all the information you need for all the books, of course, that Jeff has written, but especially hot type because that is the the latest and greatest hotness, no pun intended. AMP you AI will take you to a page where my upcoming workshop online workshop, of which we've had a few people sign up so far.

01:09:15:08 - 01:09:39:15
Unknown
It's happening on September 29th, which is a Tuesday. So end of this month, 9 to 11 a.m. Pacific. Hopefully that's a time zone kind of choice that works for most people. But go to amp ui AI, you amp AI, that'll take you to the page. And if you're interested in learning about some AI fundamentals for business professionals, well, tell me if I'm wrong here, Jason.

01:09:39:15 - 01:09:57:09
Unknown
It's also a place where you can ask the questions you've been dying to ask to understand how this stuff operates. Yeah, yeah, yeah. We have a we have a section where we'll we'll do some Q&A. So if you're confused, if you're if you're unsure, you know, Jason's a really nice guy and very patient with people, idiots like me.

01:09:57:10 - 01:10:12:22
Unknown
And he'll help you through these things and you'll have a better understanding coming out of it. So sign up, folks, I hope so. I certainly hope so. Wow, it's only three weeks away. It's like right around the corner that's going to be here before you know it. Yep. Anyways, hope to see you there. And to those of you who signed up already.

01:10:12:23 - 01:10:32:11
Unknown
Hey, we appreciate you. I'm so excited that we're going to have some people there. So let's get some more AI inside show for everything you need to know about this here show that we're doing on a weekly basis. Everything, including link out to buy some merch, which, by the way, I just have to say the quality of this shirt is excellent.

01:10:32:12 - 01:10:51:09
Unknown
I got one, I got one when I signed up for this and it's just it's super soft. The the print that they put on it is really professionally done. Like it's a, it's a high quality shirt. So I just just got to throw that out there. This is a comfy shirt but you can find the link on it AI inside dot show for the store.

01:10:51:09 - 01:11:16:05
Unknown
And then finally Patreon.com inside show is where you can go to support Jeff and I on a deeper level, and we appreciate when you do that. You get access to the daily podcast, you get access to ad free episodes of this show and discord and all sorts of things at a certain tier. We read your name like Doctor Do, Jeffrey Marikina, Radio Asheville, 102.7, Dante, Saint.

01:11:16:05 - 01:11:36:10
Unknown
James. Bono, Derek. Jason. Jason. Brady, Anthony. Downs, Marc. Starker and Carson. There we go. We got all you in there. We'd love to add another name to the list. So thank you for your support everybody. And finally, a quick thank you to Victor who is editing this one day early. I'm sorry I didn't give him advance notice this morning.

01:11:36:10 - 01:11:52:22
Unknown
I was like, oh no, but he's he's in. So you're going to get it on. On the day that we recorded it, Victor and Daniel Croft who does our social behind the scenes. So thank you, Daniel, for always being here to keep us always being here. Man. It's so good to see Daniel in the live stream each and every week.

01:11:52:22 - 01:12:09:05
Unknown
So thank you everybody for for being here and for your support. We will see you next week. Oh, and have a great trip Jeff. Thank you. Have a great tomorrow. I'm just going to Manhattan. Okay. Well, have a great trip to Manhattan. That sounds luxurious to me. I mean, I'm clear on the West Coast. It's an old media event.

01:12:09:05 - 01:12:21:16
Unknown
I'm curious to kind of see how old media are talking these days. Yes. All right, we'll have funnier event. We'll check in on that next week. And we'll see you for another episode of AI inside. AI inside. Take care everybody, by.