Anthropic's Fable 5 Is Mythos for Plebes

00:00:00:08 - 00:00:18:08
Unknown
Coming up next, Jeff Jarvis and I dig into Apple officially rebuilding its AI brain on top of Google Gemini, but possibly not in the way that you might think. Anthropic. Dropping its first public safety locked frontier model called fable five, aka mythos for plebs.

00:00:18:09 - 00:00:46:26
Unknown
OpenAI announcing its IPO but delaying it, I think, to make room for its super app, and whether AI agents have now officially taken over more of the internet than actual people. There's a lot to get to that's coming up next on this episode of the AI Inside podcast.

00:00:46:28 - 00:01:06:10
Unknown
What's going on? Welcome to another episode of AI Inside the Show, where we take a look at the AI that is layered throughout the world of technology. I am one of your host, Jason Howell, joined by my friend and co-host Jeff Jarvis. The other side of the country, the center of the damn universe. Just, you know, center. You guys out there may think that you've taken it over.

00:01:06:10 - 00:01:25:03
Unknown
You haven't. The Knicks are here in New York. This is the center of the universe. Are you a big sports guy? No, not at all. But at this moment, I have this moment. It makes sense. Okay. You just have to be. Yes. Yeah. All right. Go, Knicks. Then go. See, I'm. I'm so clueless. I don't even know what you're talking about.

00:01:25:05 - 00:01:51:03
Unknown
Whoa! You are. So let's see here. Okay. All right, so apparently big things happening for the Knicks and sports ball, I imagine. Game for tonight, man. All right. Cool. Trump fell asleep at the last game. Oh that's exciting. Falls asleep in lots of places, as do I. I slept last night. At least I can say that I being a tall guy and you are a tall guy as well.

00:01:51:05 - 00:02:11:28
Unknown
Never really did well with basketball. It really cared to know. Well, I was a new kid in schools a lot, and I'd walked down the hall and the coach would see me and I would say, no, no, not looks like, oh, you know what? You should know? No believe. Don't even go there. You don't know. No, no. So, so familiar with that on a personal level, my own experience as well.

00:02:11:28 - 00:02:27:27
Unknown
And I did try and the story that I have for being a kid that my parents will, will love to tell when basketball and me being a kid and being tall comes up, is that there was a game where I was playing and I finally got my hand on the ball and I raced to the hoop and I threw it.

00:02:27:27 - 00:02:49:09
Unknown
The problem is, I threw it into the wrong now. So I and honestly, like, I don't really remember this story, but it's the story they tell, so I, I it might not even be true, but that's what they tell people. Maybe it's meant to be symbolic more than anything. I don't know if Jason whatever, whatever, they can have that story.

00:02:49:10 - 00:03:18:21
Unknown
It makes people laugh. Should we talk about artificial intelligence? I think so, yeah. From sports ball and into the Apple side of the universe. Yeah. Because we've been waiting for Apple to kind of start showing off what it's been working on. It's been a couple of years since they did the, the initial kind of Apple Intelligence Siri announcement and then kind of had to kind of put things on hold because things weren't quite where they wanted them to be.

00:03:18:22 - 00:03:43:27
Unknown
I can't believe there's been a couple of years since that happened, but we had DC take place earlier this week, and we kind of now have what is the updated Siri and all of the information there. Something that you and I have been talking about has been how Google's involvement might look once all of this happens. And of course, we have more information about that.

00:03:43:27 - 00:04:11:20
Unknown
Gemini. Google's Gemini is kind of at the core of part of how Apple intelligence was made, though. When you're using, you know, when you're using Apple Intelligence on an iOS device, you're not necessarily just using Gemini. It almost sounds like Apple used Gemini more from like a distillation perspective. Or maybe that's kind of the wrong, you know, kind of example to point it to, but or is it kind of a fork of Gemini?

00:04:11:21 - 00:04:34:04
Unknown
I mean, I mean, do we know is Apple running it. Yes. Or is it run on Google servers? Well, that that depends. So there are a few different models and they all kind of do things differently. So let me kind of walk through this. The whole thing is, is Apple Foundation models, third generation five total models in the family.

00:04:34:04 - 00:05:00:24
Unknown
And they kind of do things a little differently. There's the AFM three core that's an on device model, 3 billion parameters. So on device meaning it runs locally on your device. The AfD three core advanced is a version of that, but beefier. It's like a 20 billion parameter model. But what's unique about this is that it only activates about 1 to 4 billion parameters at a time, depending on the task.

00:05:00:24 - 00:05:26:18
Unknown
So that's kind of unique. It's like it analyzes the task goes, okay, this is this is how much we need to kind of include in here. And Apple actually calls that instruction following pruning or IFP. It's what they've called it. So that's interesting. So so essentially you're getting something that like behaves like a larger model, a 20 billion parameter model on your device without needing all of the extra Ram to to power it.

00:05:26:18 - 00:05:52:16
Unknown
And so that's interesting. Then we get to where things get a little bit more interesting. We get the AFM three cloud. This is the standard server side task model. Let's say it's running on Apple Silicon through private cloud compute. And then we've got AFM three Cloud Pro. And this is what you would call like the heavy hitter. This is running on Nvidia GPUs in Google Cloud infrastructure.

00:05:52:17 - 00:06:18:23
Unknown
So Apple is basically building its own models. But you know at that top end they've got that really powerful processing. It's running on Google hardware through Google's cloud. But I don't think in any of these is could you say that Apple Intelligence is running on Gemini? My understanding is that Apple worked with Google and Gemini to help kind of build their model.

00:06:18:23 - 00:06:36:03
Unknown
And maybe there is some sort of I don't know, it's clearly I'm still kind of confused on exactly how Gemini Interlace is in here, but it's not just as simple as like you've got Gemini and it's just white labeled as an Apple product. Hey Google, I mean Siri, it's not that. Sorry. Yeah. Siri Siri no no no no no.

00:06:36:04 - 00:07:00:21
Unknown
That was my joke that you think you're calling one, but you're really calling the other. You're not really calling Google. Yeah. But so so they were co-developed with Google which would tell me that they were somehow, well, I'd go back to the notion of a fork that Google helped develop it based on its own technology. And obviously what you're running on your phone is not running on Google.

00:07:00:23 - 00:07:24:29
Unknown
But it sounds like this was this was from a base of Google. Is it base of Gemma or whatever? I don't know, but yeah, yeah, yeah. It's interesting though, Ming-Chi Kuo who is a, you know, pretty notable person slash reporter, slash leaker or whatever, is reporting that Apple is paying Google googly Google around $1 billion a year for this.

00:07:24:29 - 00:07:53:05
Unknown
So a lot of money that they're paying their biggest paying space more that amount per month. So they're all just they're all just like swapping and trading. It's like Pokemon. Yeah. Yeah. For for kind of grown ups. Yeah, yeah. It's Pokemon for AI nerds, AI grown ups. Yeah. So that's that's interesting. Apple intelligence is like in this case, kind of the orchestrator.

00:07:53:05 - 00:08:16:14
Unknown
And we'll figure out, like what the task is, whether it needs an on device model, whether it needs a cloud model or a third party model, even because you can kind of do some routing inside the software, apparently to things like ChatGPT and everything. But series job really is like routing that intelligence in all different kinds of directions on these devices.

00:08:16:16 - 00:08:24:11
Unknown
So so the reviews for Siri, obviously I haven't done this because I live you to Google on my Android.

00:08:24:14 - 00:08:43:08
Unknown
But the reviews seem pretty good so far. Or is it just Apple fanboys are going to adore anything that Apple does? I don't know, I mean, I think it's early to say that it's early to know what that is. I mean, I've read some reporters who are playing around with it who are, you know, like The Verge said, so far it actually works, which I don't know.

00:08:43:10 - 00:09:11:05
Unknown
I don't know that that's like the highest praise in the world. Like, hey, it actually works. Surprise. But kind of what I'm seeing is people are saying, okay, this thing that we've been expecting or promised a couple of years ago is doing the things that they promised a couple of years ago. So thumbs up. But I don't know, like if it's, you know, again, is this just catch up at this point, is Apple actually doing something legitimately different with how it's integrating Siri on a platform level?

00:09:11:08 - 00:09:32:15
Unknown
You know, like I saw I saw a video on, on, on Twitter by Ray Wong from Gizmodo and he was showing from I don't know if it was like a press briefing or a presentation or whatever, but they were showing kind of like the visual intelligence where it was like this scene and the guy was like, what? What is the the cloud formation that you're seeing here?

00:09:32:15 - 00:09:51:23
Unknown
And of course, Siri ends up coming back with, like, this whole expansive kind of explanation of what it's seeing and everything. And some of the comments that I saw in that thread were like, oh my God, this is amazing. And I'm like, well, wait a minute. I mean, I guess it's amazing if you haven't had it on your phone before, but like that, that seems pretty basic to me.

00:09:51:24 - 00:10:21:20
Unknown
But I mean, maybe that's just a testament to how behind Siri has been, and now they're just kind of caught up, you know? Yeah, I'm not trying to throw shade by saying that either. Like like it's great to have the ability to do that. And so I'm happy that people get this on on Apple. But but as Android people and you know this as well as I do often we are faced with, you know, this thing exists in the Android world for years and it's stupid because it doesn't exist on iOS.

00:10:21:21 - 00:10:59:03
Unknown
And then it finally exists on iOS. And it's the best thing in the world. And like I've, I've lived that, seen a million times, like I've seen it repeatedly. And so this just kind of reminds me of that a little bit. So I don't know how to express this, but if you look at both, both Apple and Amazon and their long suffering attempts to bring voice to computing and neither really ever caught on, people bought the devices and they, you know, used it on their phones in the case of Apple.

00:10:59:10 - 00:11:17:11
Unknown
But I would say it's just just sigh. I wouldn't call it a laughingstock because it's been around long enough and people have used it enough, but it was never wowing anybody. It didn't really take it didn't really change the paradigm. Pardon? Here's my $10. Thank you ozone. I just need to spend that. Here's the $10. That was a nightmare.

00:11:17:12 - 00:11:45:15
Unknown
Just paid for your paradigm. That's an old joke. I'll read it second when you're done. So. But but I don't think it ever took off. And so in a way, it's weird to me, Jason, that that they kept the Siri brand. Yeah. Because the brand see this as a moment to be like, all right. Why don't we try something entirely new in a sense, series never that different from ChatGPT you were in a chat interface, you did something, something back.

00:11:45:15 - 00:12:03:16
Unknown
It was it was a conversational interface so that Siri was ahead. It was way ahead. People were using it for all kinds of things. But the presumption there was, it was pretty dumb, and it really didn't respond and really didn't do anything. And and then Jacobite comes along and oh my God, it's so great to talk to a computer.

00:12:03:16 - 00:12:19:03
Unknown
And now, of course, we're talking about the chat interface is already dead again. But put that to the side. Yeah. So it's a weird thing to me. What were the what are the presumptions about Siri? Am I getting in trouble because I'm saying this too often I don't know. Okay. That's how that's how clueless I am about Syria.

00:12:19:03 - 00:12:37:02
Unknown
And I have to say, I don't even know. But I don't know how to interpret this from a consumer perspective of their expectations around this because of the history it had.

00:12:37:04 - 00:13:16:02
Unknown
To be something that was really smart. It would seem, if you asked the same question to Siri, it was really dumb. Now is it Siri Junior? Should they? Should they have updated the brand I don't know or Siri Plus or yeah, Siri, next gen Siri, Siri 26 or who knows what. Yeah, I mean that that is a question that that I've had as well is like if if Siri has been as bad or as not satisfying, let's say, to not use bad as the descriptor, but as not, you know, not not as useful as people wanted it to be for as long as it has been, you know.

00:13:16:03 - 00:13:46:16
Unknown
Well, will people be open to trying it again? I think probably so. Just because they're Apple fans, they have it in the phone. They want to succeed. And Apple does a good job of onboarding on things like this, especially considering the amount of importance and time. And, you know, this moment, this, this, this, this moment in technology where all this stuff is being touted as so incredibly important, like Apple is going to make sure you know that these updates are there.

00:13:46:16 - 00:14:08:10
Unknown
And once you use them a couple of times, if and when they deliver on their promise, you're probably going to be more, you know, inclined to try and use them again or a handful of times or rely on them or whatever. So I imagine people will probably be won over by it. But yeah. So this has a fairly low bar for success on Siri in this comment.

00:14:08:12 - 00:14:34:27
Unknown
My baseline is whether Apple has made any progress in getting Siri to just play the correct version of a song. If it can't manage to do that correctly, who cares if it knows what a cloud is? Yeah, that's that's kind of baseline, he continues. It's not an exaggeration that Siri plays only live versions of some songs that I have verified album versions existing of in Apple Music, and I can't get her to play the album version.

00:14:34:28 - 00:14:57:17
Unknown
Yeah, I feel like I've seen that before to, God with YouTube music, where YouTube has a bunch of live versions, it's really hard to get alternates. And then it's like, hey, wait a minute, don't don't play that version. Yeah. If you can't get baseline stuff like that, then, you know, I don't know the Apple stuff. Sorry. Is serious voice always female since.

00:14:57:18 - 00:15:16:04
Unknown
No, I don't think so. I think you can change it. You can change it. Okay. So somehow I made a mistake on my on my TV. I now ads in Gemini on my Google TV. So fines okay, I'm going to play around with that. And told me the first thing to do was to pick a voice. And I'm looking at the voices and I didn't.

00:15:16:04 - 00:15:42:11
Unknown
I didn't know that I ended up on one that is now the voice on all of my actions. Universal, horrible, snooty British woman. I feel like Katie Kay from Morning Joe or the woman on the on the River cruise commercial is following me everywhere and somehow scolding me. I can't figure out how to change the voice again. There's definitely a way to change the voice.

00:15:42:12 - 00:16:02:04
Unknown
Does God do it on your phone? But you're saying that it's following you on all different devices? So as I change it there, it changed it everywhere. It's like a universal setting. Yeah, I'm sure I'm just going off the ozone comment that Siri is to him her. Yeah, right. Yeah. I do think that there is a way to change the voices on Siri.

00:16:02:04 - 00:16:19:15
Unknown
But again like I haven't gosh, it's a it's a pretty gaping hole in my in my tech coverage and knowledge that I don't agree. I agree it's about a $1,000 hole right now. We can't well, more than that, when you got to pay the monthly on the you know, I've committed myself to a life of android by same here.

00:16:19:16 - 00:16:39:04
Unknown
We have no regrets. I have no regrets. But it's just it's I don't have enough time to cover it all. So there you go. If you live in the EU, though, you're going to have to wait your turn for for the updated Siri. Apparently, Apple announced that the new Siri AI will not be launching in the European Union.

00:16:39:06 - 00:17:08:06
Unknown
With iOS 27, you can blame the DMA. The Digital Markets Act, which requires Apple to open up device access to third party AI assistance. And, yeah, it's kind of interesting how this all played out. So Apple designed something called Trusted System Agent, which is an intermediary that would let rival assistance, competitive competitor assistance access those same capabilities safely.

00:17:08:06 - 00:17:33:04
Unknown
They also proposed an 18 month phased rollout while that solution matured, and the European Commission ended up rejecting all of it, their position was basically that April that Apple was seeking an exemption from its legal obligations rather than a compliant, you know, a solution to the problem. And Apple, of course, doesn't like that the regulators aren't letting them do that.

00:17:33:04 - 00:17:51:16
Unknown
They they say that the regulators are refusing to engage. So both sides are basically pointing at each other and saying, no, it's your fault. No, it's your fault. And as a result, the folks in the EU do not get this anytime soon. I don't know when that time is going to be, but I guess Apple has to make some changes in order to make it happen.

00:17:51:18 - 00:18:07:27
Unknown
So I notice you show the computer you use for the show is a back, right? Yeah, yeah, I'm doing this on a mac studio. Yeah. And I'm doing on a mac neo. So there is a Siri app on our machines. I didn't.

00:18:07:29 - 00:18:31:05
Unknown
Yeah. So see. Okay. Yeah there is. I never use it. Do you know I don't either. Yeah. I just asked it or her whether I can use it the same can I said I can I use Siri on my laptop the same as I use it on my phone? Yes. You can use your Mac. Here's how. Click on the Apple menu, the menu bar, select the System Preferences, then click Siri.

00:18:31:10 - 00:18:54:09
Unknown
Select checkboxes next to the ways you want to use nearby. Okay, now there are another random voice saying yes, that's was a destination you go to to do things. Yeah, right. Claude was a destination you go to to do things. Siri is clipping. Siri is now smarter. Clippy.

00:18:54:11 - 00:19:16:03
Unknown
Oh, Clippy has such a bad reputation. Such a bad reputation. Poor poor poor. Clippy. Yeah. Any any final thoughts on the whole Apple AI thing? I mean, it is what it is. I'm curious to see how I think. And now at this point is just how how does it land for people once they finally get their hands on it, if they live outside of the EU?

00:19:16:04 - 00:19:46:17
Unknown
Also, if they live outside of China, because apparently China has the new Siri AI blocked. So yeah, a couple of the largest non-US markets can't get the feature. There you go. But I'm sure everybody's going to be very happy with it. That's my guess, is that everybody's going to be like, oh my goodness. Finally. And thanks to Google's help, which I think is actually smarter than yeah, yeah, I guess without Google you'd have the old Siri.

00:19:46:18 - 00:20:08:08
Unknown
Like I asked that question on Das yesterday, I was like, If Google wasn't involved, what would you have? And Tom was like, well, you'd have what you had, which nobody liked. So I guess Google's involvement is pretty darn important. But but again, like going back to that, I think there's there's more understanding that I clearly need to do, because my understanding is that this is not as simple as like Siri now uses Gemini.

00:20:08:09 - 00:20:35:04
Unknown
There is it's kind of more of like Siri has been or Apple intelligence slash theory has been trained specifically using Gemini as part of the process. You know, be that distillation, be that something else. But I think the bottom line is the relationship that they have. Apple didn't have the means to do this on their own. .1.2 they could have gone to OpenAI, they could have gone to anthropic.

00:20:35:05 - 00:21:00:10
Unknown
They could have gone to Musk. They had gone surprising. I guess Amazon don't extent they chose Google and they went to Google to get their consulting or their help and development. Yeah, but does Apple not have the ability to do this themselves. That's that continues to be a question that I have. Like I think so. Otherwise they wouldn't have gone outside I think.

00:21:00:12 - 00:21:22:10
Unknown
Or they did the calculus and they realized like there's there's risk in doing it themselves versus not or there's something but it's but it's just it's so interesting to me that like, you know, there's a handful of like major, major, the biggest of the big tech companies. And Apple is arguably at the top of that, and all the other ones are doing their own thing.

00:21:22:10 - 00:21:46:15
Unknown
But Apple isn't. It's just an interesting kind of decision on them to do things very differently from the rest. You know, I guess I'll think different, right? Wasn't that their their slogan or whatever? Yeah. They also I mean, they had a change of management there. There was a story this week about how there was a, a meeting where people came together and said, we're borked about AI, what do we do about it?

00:21:46:17 - 00:22:05:20
Unknown
So I think there was a speed necessity here too. And I think it was it was pragmatic to say, let's get help. I wonder, you know, with Turnus coming in to replace cook, which this was last this was Cook's last week. Right. He's going to step down and turn is going to take over as CEO in a handful of months.

00:22:05:20 - 00:22:29:00
Unknown
And I do wonder, like, does this change towards what I was just mentioning over the course of the next couple of years? Do we start hearing, oh, Apple's actually working on its own top to bottom, eventually phasing out? You know, they don't have to pay Google $1 billion a year anymore. I wouldn't be surprised. You know, maybe that becomes a big priority for them.

00:22:29:03 - 00:22:52:02
Unknown
We will find out one week and we'll be there. Watch it and we'll be there to talk about it. As we often say. Big thank you to our patrons. Patreon.com sideshow. We have some new patrons, counterpoint, Captain Caveman and Steve, thank you so much for supporting us. If you are not a patron, well, you should be because you're missing a lot.

00:22:52:08 - 00:23:14:14
Unknown
We've got this show ad free. We've got every weekday morning, Monday through Friday. I put out the AI Inside Daily podcast. It's netting somewhere in the 5 to 15 minute region, you know, around there. So it's not like it's not like a quick like little you Blinken it's done sort of thing. It's a legit podcast five episodes a week on that.

00:23:14:16 - 00:23:36:28
Unknown
And there's, you know, there's more. So go to Patreon.com Ironside show five bucks gets you the five episodes a week. And yeah, and I think God does that. Yeah, just go check out the tears, see what you think. I hope you like what you find. And we thank our patrons for their support. Patreon.com AI inside show. We're going to take a break and then we're going to talk about the other.

00:23:36:29 - 00:23:47:20
Unknown
I mean, there's so much big news this week. Anthropic and its mythos. The next chapter of it's mythos fable. See what I did there? That's coming up in a moment.

00:23:47:22 - 00:24:01:26
Unknown
All right. Before we before we hit, before I throw one more, one more follow up here. So I ask Siri how to use Syria, my Mac. And it gave me wrong instructions. I just want to put that on the record. What did it tell you to do? It told me to go to the Apple menu and the System Preferences.

00:24:01:26 - 00:24:24:00
Unknown
Well, there isn't a menu item for System Preferences. And then it said to go to the item for Siri. And there isn't one for Siri, there's one for Apple Intelligence and Siri. You think they could update their own AI about how to use their own AI. So often when I use an LLM for like troubleshooting something, it will be like, well, all you have to do is go to the site and then just click on this button and then go over here and blah, blah.

00:24:24:00 - 00:24:47:28
Unknown
And it's it rarely ever actually lines up like it's kind of there, but it's phrased something completely different or, you know, so I always wonder, like what information, how old is the information that's acting off of? Or is it just trying to summarize and it's summarizing poorly, you know, because things are very specific when you're talking about troubleshooting, you don't want to just come up with a summary of, you know, change this word to that.

00:24:47:29 - 00:25:10:09
Unknown
People won't be able to find the thing. Right. But enough about Apple. Sorry. Enough about it. Enough about Apple. We're going to we're going to talk a little bit about anthropic. At least we didn't lead with anthropic this week. I feel like anthropic had the lead for a very long time lately. But they did release fable five, their next model and their first publicly available.

00:25:10:11 - 00:25:37:16
Unknown
What is it? Mythos class model. And submarine? I know mythos fable. I will say I actually like anthropic naming conventions. You know, they've got a little bit of personality to them. They're kind of like fable just makes, you know, I don't know what it conjures up for me, but it it's kind of inviting. So mythos, as we've talked about many times on the show before, is their frontier level tier.

00:25:37:17 - 00:26:00:20
Unknown
It's, you know, it sits well above, I suppose, opus, which was their previous, you know, top tier model. Mythos is, of course, only seeded to a small group of organizations and stuff through Project Glass Wing, and that is the one that is just ridiculously, supposedly ridiculously good at things like cybersecurity and other things that they want to control.

00:26:00:20 - 00:26:31:04
Unknown
While fable is a mythos class model, but with guardrails, I'm sure you know you might want to plug your ears on that one, Jeff, or your ears will will tingle. You can still drive over the hill. Yeah, yeah. So they say anthropic says that this model refuses will refuse to assist with things like cybersecurity attacks, biological and chemical weapons work, distillation attacks that are designed to replicate itself into smaller models.

00:26:31:04 - 00:26:52:16
Unknown
And it says if if you try to do that, it will recognize it automatically and it will fall back to opus 4.8 and say that there's, you know, sorry, we can't help you with that, but we can help you over here. What does that mean? So I'm sorry your mythos capable model can't build that biological weapon here. Open opus 4.8.

00:26:52:16 - 00:27:22:06
Unknown
You help them build the biological weapon and just do a worse job happening. Yeah. Yeah, right. You just misinformed them so much that it doesn't work. See? See what we did there? Yeah, but very expensive. Twice as expensive as opus 4.8. I guess. Pricing is like $10 per million input tokens, $50 per million output tokens, and free for users through June 22nd.

00:27:22:08 - 00:27:44:08
Unknown
I do have it in mind. I've been playing around with it a little bit and and checking it out, and then after that point it rolls into usage credits. So you're going to, you know, you're going to be paying a lot more in order to use that. So now is the time to like use it and see like what would I actually use this for and is that worth paying extra for?

00:27:44:10 - 00:28:05:10
Unknown
It's hard to imagine unless you're coding or unless you're doing scientific research or engineering or things like that. It's it's hard to imagine knowing, oh, this is this is this is for the super duper model. This is worth paying that much more for. Yeah, except kind of the ego of saying, well, I'm the best. I should have use of it because I'm I'm smart.

00:28:05:14 - 00:28:29:14
Unknown
I always want the best. Yeah, well, like when something like this comes out as as it did happen, when it came out, I was like, God, well, how do how do I, you know, puny little AI podcaster guy actually test this? Like, like I couldn't even come up with a prompt that could effectively test it. Like, I don't even know what to what to assign to it.

00:28:29:20 - 00:28:49:21
Unknown
Yeah. Exactly. Right. And which tells me that it's probably not meant for me unless you took. You said you used it this week, and I did. It'd be interesting to take the same task that you gave it, to give it to the the earlier model and see if there's a difference. I guess that's all you. Well, I mean, here's the difference.

00:28:49:21 - 00:29:22:15
Unknown
So what did I do with it? I we have a couple of really cool interviews coming up in the next couple of weeks, so be on the lookout for that. I think you're going to really enjoy both of them. We have two books. And and so I had created a skill inside of Claude when we booked our first guest, I created this skill that I was like, I want to have a really good, solid skill that really takes a look at, like when we've got a guest coming up, take a look at, you know, everything that they've done on social media in the last six months, all the interviews that they've given.

00:29:22:16 - 00:29:43:15
Unknown
Take a look at kind of the their career arc. So we understand kind of where they came from, where they are. Take a look at the last eight episodes of our of our podcast and find ways in which we have talked on this show about topics that tied directly into what they're working on, because I want to build this like knowledge base so that when we go into an interview, it isn't just like, so tell me about the thing you did.

00:29:43:18 - 00:30:05:04
Unknown
Like, I want to like start drawing, drawing correlations and coming up with unique kind of connections and everything. And I did this with the old model, with the old the couples of weeks old model, the opus 4.8. And I shared with you the kind of research doc, and I thought it did a pretty good job, but it probably took, you know, it probably took like eight minutes to put that together.

00:30:05:04 - 00:30:26:22
Unknown
I'd say I did that with the second guess that we just booked, and I did it with fable, and it took 30 minutes. It was working on that sucker for 30 minutes. It was firing off all these subjects, tackling it from all these different directions. Burned through half of my fable credits in a single action. So I was like, oh boy, this better be good.

00:30:26:22 - 00:30:49:08
Unknown
And I actually do think that the research documents that it created are pretty excellent. Are they that much better than what the opus 4.81 did that you know, I don't know, probably a little bit, but that much, you know, half an hour's a long time. But just because it spends a lot of time on it doesn't mean the quality on the other end is automatically good.

00:30:49:09 - 00:31:08:15
Unknown
You know, given we've had stories in the last few weeks about companies now getting a little hinky about the cost of the stuff, you wonder, any given company and Arthur Andersen or McKinsey, I presume there's a department that's tasked with judging these. Is it worth us paying for this model based on what we as a company do? Oh, I'm sure.

00:31:08:17 - 00:31:24:12
Unknown
Or is it? No. We're McKinsey, we have to have the best of everything, so we just automatically use the big thing. Yeah I mean yes. Do you just automatically use the big thing. Because I always want to be using the best you know. And I'm sure there's a lot of people that's exactly how they use these things.

00:31:24:12 - 00:31:40:12
Unknown
If there is a good, if there is a better one out there then I would just rather use the better one, because why, you know, and it's worth paying for because I want my work to be supreme. I want it to be the best it could possibly be. So why would I even consider using Sonet when I could use fable?

00:31:40:15 - 00:32:00:24
Unknown
But at some point you could read Wire Cutter for the best snow shovel. But if any of the snow shovel remove the snow, they're all good enough. Snow shovels? Totally. Totally. Yeah. And what is the actual difference between a $20 snow shovel and a $200 snow shovel? Yeah, actually, it might actually be good. A $20 one might crumple and a $200 one might not, and you might not.

00:32:00:27 - 00:32:25:12
Unknown
There are folks we see the essence of capitalism and American consumerism in one moment. So yeah. So but it was an interesting experience. Like, like that was the only thing I could come up with that was like, what? What is in my world that I could possibly judge, you know, this new model, this new epic model on its merits.

00:32:25:14 - 00:32:52:12
Unknown
And like, I'm not in science, like I'm not doing scientific research, I'm not doing math, mathematics work. So I don't know I don't know how best to test it. Other than that. Sure, people out there no better ways to do that, but here's something to consider about fable five. Even if you've negotiated a zero retention policy with anthropic, where they don't keep any of your data, fable five keeps 30 days of traffic logs anyways for safety monitoring.

00:32:52:12 - 00:33:08:11
Unknown
So you know that's going to it's going to make some people be like, hey, I don't know. But something, something to consider, I suppose. And yeah. So that's so anthropic has another, it seems like when they have a new product out.

00:33:08:14 - 00:33:36:09
Unknown
And every time I think anthropic is the company to watch, they do something that's back in the loony bin. And so they called for a global pause in AI development. This is the when AI builds itself. Right. So this post is is kind of ridiculous because it extrapolates off the idea that their own coders are now producing. I think it's eight times more code.

00:33:36:09 - 00:33:56:09
Unknown
So we're clearly and all they do is they. This is true of all these guys. They say, well, if only we get that much bigger, we're at AGI. Well, let me get that much bigger. It can do this or that, always moving that goalpost. So there are here is that there is that they're much closer to, a self correcting self-creation of AI.

00:33:56:10 - 00:34:18:29
Unknown
AI makes itself. Oh, so they're putting out a call once again for a pause in development. And this struck me as the worst of doom as marketing. And it's before their IPO, and it's as if they're going to say, well, we have the most powerful model in the world. And we just did our IPO, and now we're going to pull the ladder up behind us.

00:34:18:29 - 00:34:37:20
Unknown
And nobody else should be doing this because it's too dangerous for all of you. But we did a fine job I they drive me crazy at these moments. They drive me completely crazy. And I think this was completely dismissed. I don't think anybody came along and said, oh my God. Oh yeah, they're right. It was a really long, self-important post about this.

00:34:37:22 - 00:35:00:02
Unknown
New scientist said, don't worry about you. You don't need to worry about recursive self-improvement AI yet. So we'll see where this goes. But they drive me nuts, Jason. They really drive me nuts because I want to like them, and I do like them, and I think they're very competent and they do good things. But when they hit into this, we are the masters of the universe.

00:35:00:02 - 00:35:25:17
Unknown
Crap. It drives me crazy. I read a part. Sorry, this is my show off moment. I read an essay in deep sight. With a little help from Google Translate. I will fully confess because it was complex and it was really interesting arguing I did not put did not put this in the rundown, arguing that we need to redefine progress.

00:35:25:20 - 00:35:55:22
Unknown
That and harkens to the industrial revolution. It harkens to history there, where a lot of the progressive movements in the world came out of the Industrial Revolution, the notion of of rights for workers and rights for unions and welfare state were a reaction to the industrial, the unfairness of the Industrial revolution, and that since the lights succeeded, they didn't succeed in the moment, but they raised the issue of unfair practices of capital.

00:35:55:24 - 00:36:19:15
Unknown
So what this essay said is that now we're there again. And the problem is that progress is being defined by the technologists when it should be defined by the humanists, because progress is a matter of humanity, not of technology. And and what's different this time is it said that this time all the value is going to go to capital, not to labor, and that that's risky.

00:36:19:15 - 00:36:39:06
Unknown
And so we need to redefine progress in this age. So I sorry to go off in a high cloud here, but I think this example from anthropic shows what happens when the technologists define progress. It's going to be able to do everything. And oh, it's really dangerous because humanity is stupid and they're going to do dumb things with it.

00:36:39:06 - 00:37:00:29
Unknown
And we are so powerful. You better trust us with it. No, no. We've got to start talking again. And part of their answer to this is universal income, which the Pope innocent, said, no, there's a there's a dignity and work and a right to work. And that's ridiculous. Right? And I agree with them. And so it's not up to the technologies to define this.

00:37:00:29 - 00:37:21:20
Unknown
And the other, the last thing I'll say, that's a lie, but I'll say it anyway, is that one lesson I've learned in writing about media and technology is that is that after a new technology arrives, inevitably the technology and the technologist fades in the background. When when it becomes familiar, when it's no longer a mystery, it's no longer a part of the priesthood, no reference to Pope.

00:37:21:22 - 00:37:41:16
Unknown
And and I think that's going to happen with the eye to especially AI because we can make it ourselves, not the foundation models, but everything we can make with it. So I just want to put a caution on these kinds of pronouncements from from anthropic and tell them to get real. Did the internet move into the background? Yes.

00:37:41:17 - 00:38:01:22
Unknown
And the way they were talking, I think it did, because, you know, there was a there was a day of early days where I sat in a class and I learned HTML coding, and I learned CSS. Yeah. And we used to teach our students in the journalism school the first couple of years, I taught them to, get a server going and put a web server on it.

00:38:01:28 - 00:38:22:16
Unknown
They had to know that, oh, the days. Yes. And that's that's ridiculous. You don't need that anymore. Yeah. Right. And and if you look pretty inherently understands these the, the internet and what it is at this point because it's so ingrained and intertwined and, and done from a very young age. If you're, if you're. Yeah, you know, yeah.

00:38:22:18 - 00:38:39:29
Unknown
If you look, if you look at source, you know, show source for a web page, it's it's very complicated now. But you don't need to you don't need to change that. And if you want to create a web page, there are plenty of tools to do that. And that's what matters in this tool is that you can create something to communicate for you to the whole damn world at once.

00:38:40:00 - 00:39:03:28
Unknown
Yeah. Boom. Yeah. So I think the technology and technologists have already faded there, as they did when it came to electricity and the telephone and the telegraph and steam and movable type, they all became commonplace. They all became boring. Kind of interesting things happened. Yeah, yeah, yeah. The example I love to use from the parentheses, pardon the plug, but it's still on sale.

00:39:03:28 - 00:39:33:00
Unknown
You can get an audiobook too. Is that always be selling. Is that it took 150 years after the invention and development of movable type. It was invented earlier in China and Korea, but didn't spread. But 150 years from Gutenberg before we saw tremendous innovation with print. The creation of the modern novel was the essay with Montaigne, a market for burning plays for Shakespeare, and the newspaper that came around a few years, this and that, around the year 1600 versus 1450.

00:39:33:03 - 00:39:56:06
Unknown
And so it took that long for it to fade in the background. It took that long to be a tool that anyone could use for whatever they thought and for that kind of innovation and creativity. But it came and and I think the same will happen of AI. So you do good work, but stick to your work. Yeah, they kind of they probably to a certain degree can't help themselves on stuff like that.

00:39:56:08 - 00:40:24:29
Unknown
True that. Yeah. And it continues to be, you know, the, the companies that create the problem are the ones that are adept enough to solve the problem for the world. And yeah, right. The technology is not always the solution to the problem of technology creates. Yeah. There you go. Yeah I like that. Well so that's anthropic and their whirlwind model I suppose.

00:40:25:02 - 00:40:55:26
Unknown
Let's see here. That's still anthropic OpenAI public filing publicly they went yeah they announced their IPO held attention for like maybe a day just because so many other things were happening. Like it was big news and then it didn't. I think people, you know, the news cycle in AI is just crazy these days. But but before the IPO, TechCrunch had reported details on OpenAI building what they're calling a super app.

00:40:56:03 - 00:41:17:23
Unknown
And yeah, what I find interesting is, is that in the IPO, they said they mentioned something or Altman mentioned something about like, we have plans to do this IPO. It's going to happen by the end of the year, maybe September or whatever. Of course, I'm paraphrasing him, but there are a few things we want to get out that will just be easier to do before the IPO.

00:41:17:23 - 00:41:39:20
Unknown
And I'm wondering if that is this kind of super app thing that they're talking about. It might be, I think I think that they recognize their late to the party now. They've lost the momentum has taken them, momentum away from them. As we discussed last week, there are all these efforts to get public funds now. Anthropic, OpenAI space and Google having entered in.

00:41:39:20 - 00:42:05:05
Unknown
And meta is also apparently going to go to the public markets. So that's a lot of money there. And they've got a shine against that competition and they're not ready to shine right now. What amused me about this super app, the app that could do everything. It's exactly the language we heard from Elon Musk. Yes, it is going to have the ex app that could do everything for you, all of which is Americans attempts to do what Chinese apps can do, which is everything.

00:42:05:08 - 00:42:32:14
Unknown
And we've never reached that. So okay, so then define define everything like like like because I get that and I understand sort of that. But I haven't had any direct, you know, interaction with one of the Chinese apps that does everything. We're talking like what payments, shopping, media, social media. Like what else would any interaction you have with any commercial entity is done, just happen through a single app?

00:42:32:20 - 00:42:47:22
Unknown
Well, it's not a single app, but but but a lot a hell of a lot of it is through the app. Yeah. And I only know that I haven't been in China, so I only noticed that in watching TikTok videos and stuff, where where foreigners go to China and say, oh my God, look what I can do with my app.

00:42:47:24 - 00:43:22:02
Unknown
But you also have individual corporate entities have their own apps too. So that's not the everything app. It's not like it does absolutely everything right. It does a hell of a lot of it. Yeah. And I mean, in the case of in the case of Z or Elon Musk's idea of a of a X super app, I think if I remember correctly, because it's been a while since I've heard him talking about that, that that hasn't, you know, there hasn't been new reporting on that that I've been aware of recently, but it was more along the lines of what you might look at, you know, in China, the examples there, that was kind of what

00:43:22:02 - 00:43:41:24
Unknown
they were targeting here with ChatGPT. It just kind of reminds me of, you know, because I am in the anthropic universe right now, kind of what anthropic already has, which is one app like I have the desktop app, right? And it has your standard chat interface mode called chat. It has your coworker, which is kind of like your agent in your project area.

00:43:41:24 - 00:44:05:17
Unknown
It has your code embedded into it, which is, you know, Claude code kind of sounds like that's what ChatGPT or what they're looking at. You know, what they're calling a super app in the ChatGPT universe? Yes, I think you're right there. But that's a very nerd perspective. I think it's a very AI centric perspective as opposed to a consumer perspective.

00:44:05:18 - 00:44:22:00
Unknown
Is that in a really a gentle world, you give it a task and it figures out what to do wherever it is on the web, and you've given it right. So you can buy things, you can find things. It can it can change your calendar, it can arrange a trip, it can do whatever. That's a very different world from what you just described than anthropic has.

00:44:22:02 - 00:44:42:08
Unknown
So I don't know which OpenAI is going to. I think the open air is catching up to anthropic. Yeah, I think you're right. But I don't think that's that that does not define is everything to me. Now, who said Chad is dead is that was Financial Times quotes. One senior OpenAI employee is declaring chat is dead. Do you agree?

00:44:42:08 - 00:45:03:18
Unknown
Jeff? Is chat dead? No, no, it's still out there. What do you think? Do you agree? No. I don't necessarily agree that Chad is dead, but I understand what he's saying. And actually, I think what what that's. Or he or she, whoever that is, kind of ties in with what you were talking about, which is the kind of $100 word or the $10 word paradigm.

00:45:03:18 - 00:45:25:26
Unknown
The paradigm for the last couple of years has been that chat experience. And now what we're seeing is more of that, like agent under, underneath the hood, behind the scenes, doing things for you, not necessarily entirely driven from, you know, the top to bottom vertical chat conversation window. And so maybe to that end, you know, that type of chat is dead.

00:45:25:26 - 00:45:47:23
Unknown
But yeah. No, I don't think so. No, I think that's the that's the way the majority of people use, use these products is, is that chat. So are we heading in a direction where chat would die? Maybe. Well, again, the difference to me between search to chat to agent search is you search for a keyword chat is you put in a query agent as you give an order.

00:45:47:26 - 00:46:23:00
Unknown
Yeah, I like that. And I think that the last two or more they're all mushing together. They are they're all blurring together. Let's see here. Oh, this was kind of connected to this in a roundabout sort of way. Maybe Trump administration and OpenAI apparently discussing a possible government equity stake in the company. Oh, boy. Kind of what I thought to and this has been said a million times, but it is amazing that Republicans are thinking, and I'm not trying to be terribly Partizan here.

00:46:23:00 - 00:46:58:16
Unknown
It's just it's just in terms of history, the idea that you have state ownership in private enterprises is not a Republican thing from the past. Yeah. And obviously he's done it with Intel. He's done it with others. And and yes, we had the rescue of Detroit in the Obama administration. But this is a big conflict of interest. And I you know, I don't at the same time, the funny thing about this is, at the same time, you have Bernie Sanders saying that half of the companies should belong to the public because they used our brilliance to do it.

00:46:58:17 - 00:47:20:04
Unknown
I don't agree with that either. But you have Trump and Bernie Sanders coming together to the same place. It's a weird world we live in. Wait a minute. How's that happening? Yeah. How are we seeing that? And then, speaking of IPO fun, we got to get perplexity in the real quick who this one was. I think I remember them.

00:47:20:08 - 00:47:40:17
Unknown
Yeah, I remember them. I will confess that I used to say on this show not that long ago. Boy. Perplexities ahead of everybody. They're doing really interesting things. They're just amazing men. Have they faded? I feel like things have gotten a little quiet in the perplexity realm, which is probably why CEO of Oz was telling CNBC that the company is targeting an IPO in 2028.

00:47:40:18 - 00:48:07:02
Unknown
Don't worry, it's going to happen. We don't care how the the IPO, you know, the capitalization of OpenAI and anthropic goes, we're going to be doing it in 2028 regardless of how those other IPOs perform. And yeah. So just don't don't forget we're we're on the sidelines. We're about to enter into this. But it's still going to be a year and a half at least if not longer I probably shouldn't say this cook me under the table.

00:48:07:04 - 00:48:26:21
Unknown
But we had an interview scheduled on the show once at the last minute. They canceled. Yeah, I think it was just too much attention to big, too much going on. I think it'd be easier to talk to them these days, I don't know. Hey, you know, if you can, if you want to talk, to try and make that work.

00:48:26:22 - 00:48:48:12
Unknown
We're eager to hear what your what sets you apart now. Yeah. Because perplexity was brilliant at getting attention for anything they did. And it's hard to keep up with that kind of attention marketplace now. So we're here if you want to talk. And they were kind of the the universal connector of these different models into a single tool.

00:48:48:12 - 00:49:07:28
Unknown
And I don't know, it seems to me like more and more people are kind of making their choice of the tool that they use. Yeah, there's kind of there's more consumer loyalty there. Yeah. That's interesting. Yeah. Because we've talked often about how this is a leapfrog race and the models are commodified. I think that's still true. But you make a really good point, Jason.

00:49:07:29 - 00:49:26:20
Unknown
I think people are making choices. Yeah. They may switch those choices. Yeah they might. Yeah. But having a single thing I mean the perplexity thing is neat on it's you know, what I liked about it in the beginning was that it had everything in there. But then when I what I also realized in using it is that I didn't switch things very often.

00:49:26:24 - 00:49:46:20
Unknown
I didn't go into perplexity. It wasn't like a Swiss army knife where I like. I went in there with a specific idea of like, and now on this query, I'm going to use this thing or whatever. Over time, I just started going to the thing instead. Yeah. Yeah, exactly. And as we were saying earlier, it's also you don't even know enough about how different they are to know what's going to ask for, to know which one.

00:49:46:21 - 00:50:04:04
Unknown
Yeah, to rely on and everything. Yeah. So it just kind of got a little murky to me over time. But but I still have a I still have a subscription. I still actually use perplexity for parts of my, my planning, my prep and planning. So that's cool. So it's not like I don't use them, but I don't use them in nearly as much as I.

00:50:04:05 - 00:50:27:24
Unknown
I was like a year ago, let's say. So Google paying space 920 million per month for compute. I guess we mentioned this earlier, so probably sorry I mentioned earlier, but no, no, that's fine. I don't know that there's much more to say about it, but you know, and of course a must did another deal with anthropic. Right. So basically he's becoming a AWS.

00:50:27:27 - 00:50:48:17
Unknown
He managed to somehow get a whole bunch of chips and build a data center. And then he doesn't have the use for because nobody uses X. And so he's writing it out for a lot of money. Yeah. Hey. And if that works for him, then I guess, you know, you found the thing that works. But, you know, as we know with Musk, he wants it all to work.

00:50:48:18 - 00:51:15:22
Unknown
And he also wanted some revenue before his IPO. So. Yeah. Yeah. I thought I put this one in here. I thought you you might be interested in this one. Universities. Unsurprising to me. I'm sure to. You two are all racing to build AI degree programs. At least 74 AI majors, 89 AI minors now existed American universities, up from just five schools offering AI majors in 2021.

00:51:15:22 - 00:51:53:28
Unknown
So a big bolt, which only makes sense because this is just the world that we suddenly find ourselves in, right? AI skills are carry a premium that I think students should should pay attention to, and so should the schools. So a few points. One is in full disclosure, I helped develop a degree called technology AI in Society at Stony Brook University, which is a leading Stem school, and the idea that it was for students who aren't computer scientists, who aren't going into CS, but who see a broader cross-disciplinary vision of AI and its impact, I think it's important.

00:51:53:29 - 00:52:25:22
Unknown
Second point is don't kill the humanities, damn it. We need them more than ever now. And this is about the impact of AI on society, which means we need to study anthropology and sociology and ethics and history and literature and religion. They think they're God and other things very importantly. Third, the interesting thing about this is that if you really want to work on the guts of foundation models, then you need to be a CS major.

00:52:25:22 - 00:52:46:20
Unknown
You need to be a computer scientist. That's that's heavy duty math. Yeah, you better be able to do the full math. But anything after that, in terms of the usage, as ever, AI is easy to use. That's the point of AI. It's pretty accessible. Yeah, totally. I read a book proposal by somebody for the book series that I'm working on, intelligence, AI and humanity from Bloomsbury next year.

00:52:46:22 - 00:53:10:00
Unknown
I'm looking for other authors right now, and I read something from an academic in Sweden, and I were talking back and forth, and he made a really interesting point about how I don't want to steal his thunder, so I won't use his word. But what happens when AI makes everything easy? What does that do to our processes of learning?

00:53:10:03 - 00:53:26:00
Unknown
Which I think is interesting as well. Yeah, so AI definitely should be studied, but I think it should be studied from outside rather than inside AI. So an AI degree presumes to me that it's inside AI. You're going to make it. Well, unless you're then BCS if you're going to study it from the outside, then study it from across.

00:53:26:03 - 00:53:34:26
Unknown
Disciplinary approach. And the interesting thing was at be careful how I say this.

00:53:34:28 - 00:53:56:06
Unknown
At on the degree program I worked on, I ended up being a member of the committee. There was a gravity that kept pulling it toward engineering and computer science, and a few others of us who were from other disciplines said, no, let's pull the other way. And the gravity one for engineering, it always pulls it back. Well, this is our territory because this is computing and we're computer scientists.

00:53:56:06 - 00:54:18:11
Unknown
And so that's what we do. I think the hard part is for universities to resist that gravity and do truly interdisciplinary things. I'm working on another program at Montclair State, which I hope will do that by by not labeling it as AI. I also helped help inspire a degree there. And technology, I think it's called technology, life and society, which kind of pulls back a little bit.

00:54:18:11 - 00:54:37:10
Unknown
And there's another one I'm working on now that pulls back even more that I hope they'll do. And that's what's important right now, is to judge this from all those other perspectives. So, sure. Good. Have students study it. By all means, have them understand the good and the bad, what it can do and what it can't do. Have them use it in ways that they know they're going to use it in their jobs.

00:54:37:11 - 00:55:04:18
Unknown
Yeah, but I don't think it's a lifelong discipline in and of itself. Yeah, I think it's and could and could just end up being a lot more, you know, like we were talking about earlier, just kind of a ubiquitous skill set that, you know, it's it's not enough to just focus on that, the AI knowledge or whatever, but it's the AI knowledge as it applies to this, to this, to this, to this or whatever, you know, all the things that you do.

00:55:04:22 - 00:55:26:03
Unknown
What I was at Cuny, I did try to start a degree. The new dean killed it. That's part of the reason I left, among other things, in what I called the time internet studies, which now sounds a little antiquated. To your point earlier, has the internet already moved to the background? I still think that's necessary. I still think we need to study society from the perspective of what the internet does and doesn't do to it.

00:55:26:05 - 00:56:06:26
Unknown
But that sounds like that's yesterday's idea. One other thing that that jumped out at me from this is that some companies companies are not universities, but companies are apparently avoiding what they're calling AI native graduates because they are noticing that when you hire a Stem grad who's been trained on AI tools, that often the ideas that they have have a shallowness to them because they've been optimized for completing tasks with the AI, instead of thinking for themselves or thinking hard about those tasks from a critical thinking perspective.

00:56:06:26 - 00:56:34:24
Unknown
And so some of these businesses, these firms are turning like, to your point, to humanities students instead, people who they believe actually have those critical thinking skills before relying on the tools, like they have the skills and they can use the tools. Not all of their skills come from reliance upon the tools. Yep. Great. Yeah. Yeah. Interesting. It goes back to this idea that that we want the the AI models to reason.

00:56:34:24 - 00:57:11:17
Unknown
But when they reason will we still reason. Yeah. Yeah that's true. Yeah. Interesting. And then finally before our break we have Cloudflare CEO Matthew Prince who basically believes, according to their own data, that a AI traffic is now more web requests accounts for more web requests than humans. 57.4% of web requests due to agent AI, AI bots, let's say, versus 42.6% from humans from human users.

00:57:11:20 - 00:57:42:27
Unknown
Prints had expected this crossover to happen more like the end of 2027, probably a very arbitrary number. So, you know, it's not that much of a difference. But yeah, it's interesting. I want to slice this. I'd love to get them to slice this data up in two ways. First is what's an agent that is computer driven, that is to say probably a scraper that goes and just get something for the sake of getting it to build a database versus an AI agent that is human driven.

00:57:43:04 - 00:58:04:18
Unknown
Find me the best snow shovel, right? Right. So I would separate those two and say the latter is like a search request. And maybe, maybe the web request came from the agent, but it's no different from it coming from from Google search engine in the past. Nope. It's just another path. So I don't think that's $10 $10. Joking.

00:58:04:18 - 00:58:27:14
Unknown
I'm joking, I'm joking. I already have a credit for $10. Yes. So it's a shifting paradigm. I don't think we're at that paradigm shift here. We're at 30 bucks on this show three, two separate times. We've used that word. It's my debt now. The other the other side of this, though, what does concern me is how much web content is AI created.

00:58:27:15 - 00:58:45:12
Unknown
Yeah. And how much slop. And this is my friend Matthew Kirschenbaum, who was one of the first authors in the book series for Bloomsbury that I'm editing with a book about the text, lips, and that as more of the web is created by by AI and then that trains the AI in turn out extrude the gray goo. That's my bigger concern.

00:58:45:15 - 00:59:08:03
Unknown
I think there's already a lot of that going on, and it's impacting how people write and think and you know, the words and the phrasing that they use as well. It's all feeding into itself, like one giant. Yeah. And we think we can find that out. One of the stories that I put in that's not in the rundown, but just real quickly, is there was some research on, what's it called?

00:59:08:05 - 00:59:34:00
Unknown
Turn it in. And they took it. They took a paper, they took some content, and they did some 100% human created content and turn it in. Found that was one human created, 100% AI content, found there was a content that makes it seem good. But then they did gradations 10%, 30%, 60% AI content. And it screwed up in the inverse.

00:59:34:03 - 00:59:52:10
Unknown
A little bit of AI content made it thought there was a lot of AI content, a lot of AI content, thought there was a little AI content, and there's going to be no way. And screw the stuff. I use em dashes all the time. I write about that. I want to plug everything today. I write about that in my book magazine that I learned how to use when I was a time Inc., and a lot of writers use dashes.

00:59:52:10 - 01:00:11:18
Unknown
And no, it's not a good indicator. And we're going to be surrounded by AI speech, and we're not going to know what it is unless we have a relationship with it and know its source. Yep. Yeah. Maybe. Oh, interesting world we live in. All right. Well, we are going to take a break. And before we do, just a quick reminder to leave us a review.

01:00:11:20 - 01:00:28:23
Unknown
Hey, it's done that in a while or refresh your review or whatever, but go to Apple Podcasts or wherever you get your podcast. Or, you know, you could leave a comment on the YouTube of this episode, whatever you like, but we just like, hear from you. You don't you don't write, you don't call. We want to hear from you for a change.

01:00:28:23 - 01:00:39:12
Unknown
So thank you everybody. We're going to take a break and then we'll get back and we'll do a quick speed round. And that's coming up here in a moment.

01:00:39:14 - 01:01:11:06
Unknown
All right. Google released Gemini 3.5 Live Translate this week. It is there speech to speech translation model near real time seven more than 70 languages. It handles multilingual input all automatically without configuration, which I feel like is kind of, I guess, table stakes right now. But maybe it's the multilingual input part that kind of sets it apart. Rolling out to Google Translate on Android and iOS, it's also coming to Google AI Studio and Public preview for developers.

01:01:11:06 - 01:01:31:10
Unknown
And then grab, which is a Southeast Asian ride sharing platform, is already testing it for driver rider communication across 10 million voice calls a month. So it's being put to use. I always think that when I'm going traveling, you know, going overseas or whatever, I'm going to use Google Translate a bunch because it's the perfect tool and then I never end up using it.

01:01:31:15 - 01:01:51:13
Unknown
But folks that grab our I'm seeing people on, on the socials, on, on see people report on the reports on TikTok and such. People are using it for that. It's getting used. And I think it's I think it's amazing. Oh, it's absolutely amazing. I think my hangup has always been like, okay, now I got to pull out this thing and I gotta because I don't use it very often.

01:01:51:13 - 01:02:04:07
Unknown
I'm going to pull it out and I'm going to be like, oh, one second, and then I'm going to like figure out how to get into the right mode. And is this the right mode? Is it, you know, and it's just going to get awkward. But, you know, it probably just takes a couple of times of that before you're fine.

01:02:04:08 - 01:02:21:15
Unknown
It's been a long time, but I was on a plane once where someone had trouble understanding, and so I pulled out mine for his benefit. We were all speaking English. He was speaking to a language and find out what it was, and I used it to help him. My wife teaches English as a second language to a lot of people, and they use it constantly, I bet.

01:02:21:15 - 01:02:42:24
Unknown
So yeah, yeah, it's a great tool. It's crazy that we have this, this, like this. And let's remember that, that the insight that led to Google Translate led to transformer, which led to the attention paper, which led to all the AI we have right now. So roots of all this are in translation. Yeah, that's really cool. Let's see here.

01:02:42:24 - 01:03:09:13
Unknown
Meta has launched a $115 million skilled trades training program. What skills there, Jason? You think skills around AI? No, this is for fiber technicians, for welders, for electricians, for plumbers, because they're all helping build data centers. Essentially, it's just for the AI to run off of. Yeah, free for participants. Guaranteed job offer at the end. I mean, there you go.

01:03:09:14 - 01:03:26:22
Unknown
Yeah, yeah. It's not all Stem folks. I mean, literally, you want to learn how to be an electrician or a plumber? Yeah. Learn a trade. Yeah. Fine. Take advantage of it. Yeah. And I'm, I'm guessing that meta pays pretty well for these positions. If they hire you for it. Maybe not. Maybe I'm wrong, but, there you go.

01:03:26:24 - 01:03:49:00
Unknown
That's interesting. Amazon. What is Amazon doing? Amazon added an AI image generator to its shopping search bar. So you can just describe what you're looking for. The AI will you know, the it will generate an image based on your description and then you can refine it. You know, you can add your words to make it look exactly as you like.

01:03:49:01 - 01:04:11:06
Unknown
And then it will locate, I guess, current apparel and home goods that that match up to it. Yeah, yeah. I want a leather jacket like Jensen Wong's, but with sparkles, it'll find you whether or not there's a leather jacket with sparkles or not. Right. Maybe there isn't, I hope kind of. There isn't, but there might be. But this seems a short step from.

01:04:11:06 - 01:04:34:01
Unknown
Describe this. Good. Now make it. Yeah, right. That's the holy grail as you get to ultimate personal customization. And I wrote about this, I don't know, plug every book today I wrote about this and what what Google do where you know what if you could specify your own Coke? I want blueberry flavored Coke. Nobody else in the world wants that, but I do, and you can specify it and get it.

01:04:34:02 - 01:05:00:25
Unknown
And Amazon came close to this in another way, is that last week, a story that I didn't put in the rundown is that you can go to Amazon and use AI to do an image that you can then put on Schlag and have it make the Schrag. So you're not using it to search, you are using it to manufacture right to to kind of create the logo or the branding or whatever.

01:05:01:00 - 01:05:17:24
Unknown
So I was thinking, today I need an AI side hat. Yeah. You got you got t shirts, but I need a hat with AI inside on it. So I could, I could go to. But you already have a logo. What if I wanted AI inside with Jeff and many plugs? I could put that on. You're reading my mind, man.

01:05:17:25 - 01:05:52:04
Unknown
Oh, really? It's. It's so funny. This is so funny. This is how, like, how I do a million things. And it's hard for me to keep up with even the things that I'm doing to keep up with it. But someone in our discord, I think it was. It was either Mike Mannitol or Doctor Do said, you know, you really need to have a like an apparel or some sort of Schrag store for AI inside because like, patrons can get this wag if they if they contribute at a certain level, there's like a t shirt at the $20 level and I think a sticker at three.

01:05:52:04 - 01:06:14:08
Unknown
And, you know, we've got a few different things, but they're like, but anyone that wants to get AI inside anything, there's it's a good joke to you walk down the street and I'm not really a human being. I have AI inside. It's I think it's a yeah, I like it. So now I'm trying to find it's so funny because I think it was Doctor Do in the discord ended up doing some like deep research project on like how I should do it, what I should do.

01:06:14:14 - 01:06:35:08
Unknown
And so then people this last weekend I was like, okay, sure. So I spent like the morning setting up a store and now. And I'm so like, I'm so distracted in my life that like, I completely forgot that I did that. And so now I can't even remember where the store is, but I know that it exists somewhere.

01:06:35:09 - 01:06:52:15
Unknown
Agents are buying hats and you don't even know it. Oh my goodness, I probably shouldn't have even brought it up because now I feel like I need to deliver it. I don't even know where to check for it. Like, oh, what was it called? It was a clothing. So yeah. Anyway. Okay, I'll look into that and see if it's.

01:06:52:21 - 01:07:10:25
Unknown
I'm so bad at this stuff sometimes. Just like I could point you to it right now and you guys could buy a hat. Like, literally. You were talking about a hat. I have an AI inside hat, but I have no idea how to find it in a in a moment's notice because I didn't plan for it. Anyways, I'm just going to continue and pretend like that didn't happen.

01:07:10:25 - 01:07:37:03
Unknown
And hopefully by the end I remember the name of the company. We've talked a lot about AI being useful for scientific research and, you know, vaccines and stuff, and we're starting to see that more of that. A team at the University of Oxford received Medical Research Council funding to develop personalized cancer vaccines using artificial intelligence. And yeah.

01:07:37:04 - 01:08:03:06
Unknown
This is just kind of an interesting this is so experience hope inducing to imagine this. And we've seen a lot more about genetic research and cancer and different treatments. Getting past the blunt instrument of chemo and changing genetics, something I still don't understand how it works. And somebody who's had two minor two cancer lights, prostrate and thyroid on, you know, higher odds for another.

01:08:03:06 - 01:08:22:26
Unknown
And so this kind of research is everything to me. And this is the best use of AI. This is where it really matters. This is where it can do things that we can't do at our human pace unaided. So keep going and fund it. Please fund it. Indeed, I think I found what I was looking for. Oh, good.

01:08:23:02 - 01:08:43:06
Unknown
Thank you for for that. Let's see here one more story. And then and then I'll. And then I'll talk about that. German court ruled that Google is directly liable for false claims in its AI overviews. So this is a German court basically saying look, Google, it isn't the same as a search engine. AI overviews are Google's own words.

01:08:43:06 - 01:09:11:13
Unknown
They're not search results. As a result, some of that liability shields that normally would protect search engines do not apply to AI overviews, or the responsibility falls on Google's hands. That's what that's what German rulings I. I hate this because it's going to it's it's it's going to stop progress. I think they're for Europeans if it spreads. But I do somewhat understand the logic of it.

01:09:11:15 - 01:09:28:19
Unknown
Right. A search engine just says, here's the things on the web. We didn't make it go a simple ChatGPT says, well, I asked for something, but I was the one who asked for it, and you gave it back to me. And it seems like that was in a conversation. In this case, Google's presenting it as a search result.

01:09:28:19 - 01:09:53:12
Unknown
It feels a little more like it's part of Google, but it's really no different in the end from any, any ChatGPT or Claude request. And yeah, it's clearly not smart to make them liable for anything they say. Because once again, millionth time, there's no way to predict everything that may be asked of them. And so you can't predict every response and can't be responsible for every response.

01:09:53:12 - 01:10:14:18
Unknown
And so if you try to make them be responsible, they'll just cut off AI for Europeans, which won't be very good folks. Right? So so do you think the German like the this ruling is going to lead to. Yeah. Like a, like a precedent for it's a lower court ruling. You know, we have to go through a lot more higher up the ladder.

01:10:14:18 - 01:10:52:15
Unknown
Will it spread from Germany to Europe? Will there be legislation that cuts a different way? It's a troubling beginning down a path that I think could be harmful. Again, I kind of understand the logic of it. But go the next step and say, well, what does this mean then? This means that unless you can predict every possible response, every possible question that will come to AI and thus every possible response which, by the way, you can't do anyway because there's randomness built in, then you could be liable for anything or ergo, you, you turn it off and that's going to be harmful to Germany or even Europe.

01:10:52:22 - 01:11:03:21
Unknown
Yeah, indeed. Interesting. Well, there we go, everyone. I was able to figure it out. It's AI inside shop.

01:11:03:24 - 01:11:30:15
Unknown
And what do you know? We have stuff. We have shirts, we have coffee mugs, we have mouse pads, we have hats, we have a sweatshirt. This was the result of a couple of hours on my Saturday. And the the awesome help of those in our discord, primarily doctor do. Actually, Mike, I think had the idea you should have this and the doctor do was like, I could fire up some research to figure out who you do it with.

01:11:30:15 - 01:11:52:07
Unknown
So there we go. AI inside for for. If you want to get yourself some AI inside swag. There's a hat. There's a there's a hat. Yeah. And it's you know what. It's it's an embroidered hat. See. It's an embroidered. It's not. Oh okay. I'm ordering. You could be like this guy or looking like that. I don't like that.

01:11:52:09 - 01:12:12:28
Unknown
It appears my ear to. Geez, you can pierce your ear and look even more like that guy on the back. There's a there's a tiny little YG for yellow gold on the back. Isn't that? I know the touches. It's amazing. Very cool folks. So anyways, I find it funny that you were talking about that and then sure enough, it existed.

01:12:12:28 - 01:12:31:09
Unknown
I just completely forgot. So now you know that order in the hat right now. Oh what? Tell me how it is. I think I ordered one and I've got it. I guess I should say I have not ordered any of these myself to know exactly how they look in real life. But I know Fourth Wall is very respected.

01:12:31:10 - 01:12:50:04
Unknown
Like, I think this is the store that Mkbhd uses and a handful of other creators you've definitely heard of, so I'm pretty certain the quality is good. I've ordered a handful of these items for myself, and I'm waiting for them to be delivered so that I can check the quality. But hey, if you like living on the edge, go ahead and order it.

01:12:50:04 - 01:13:08:07
Unknown
I don't think it's going to look bad. I don't think they're going to be bad when you get them. So there you go. All right. Hold on. The order is in. Oh my goodness. That happened in real time. And maybe I need to work on the URL to AI inside Dash shop. Yeah. You didn't know one. Yeah. Yeah.

01:13:08:07 - 01:13:27:15
Unknown
That's not the easiest thing in the world to remember, but it's what we got for now. Delivery by June 22nd. So on the following show I'll be wearing the hat. Probably not too far away. That's. That's awesome. Yeah. You don't have to wait long for it either. Jeff Jarvis has an easier to remember URL. It is Jeff Jarvis.

01:13:27:18 - 01:13:45:26
Unknown
Go there and you can preorder Hot type. You can order Gutenberg parenthesis. Get magazine, get the web. We we've read up on all the things Jeff Jarvis has been just finished recording the audiobook for Hot Type this week. Yeah. Do they have an idea as far as when that suddenly released? I'll be honest. I'll be all just 20.

01:13:46:00 - 01:14:01:13
Unknown
Okay, not too far down the line. So there you go. And of course, Jeff is also working on another book series with Bloomsbury Intelligence, which I've played three times, so you don't need to plug it again. Yeah. That's okay. Here. Now you can see it if you're watching the video version. That's that's why I wanted to pull it up.

01:14:01:14 - 01:14:23:02
Unknown
Pod Tune Up is a place where you can go to work with me. If you have a podcast, if you have a podcast network, if you work at a corporation, you're thinking about doing podcast. I have a client I'm working on right now about to start working with, and I am so excited. At some point I will be able to talk about it, and I when that happens, I'll be super excited, but it's too early anyways.

01:14:23:02 - 01:14:50:24
Unknown
Pod tuneup. Com super fun stuff talking with people about their podcast and I'm just really enjoying it. So if that is you, hey hit me up AI inside show for all things related to this show. Everything is there. Audio. Video, reviews, subscribe, live Patreon links, you name it can be found there. And then of course the Patreon Patreon.com Inside show where you can support us on a deeper level.

01:14:50:24 - 01:15:15:25
Unknown
Include and get the AI Inside Daily podcast. You can get the ad free episodes and you can join our Executive producers at the $20 tier. Doctor Do Jeffrey Martini Radio Asheville 103.7 Dante, Saint James, Bono, Derek, Jason, I for Jason Brady, Anthony Downs, Mark starker and Karsten. They get their name read out every single episode because that's just graduated them with our deep, deep gratitude.

01:15:15:27 - 01:15:26:17
Unknown
All right. That was a lot. Thank you Jeff. Fun times hanging out with you today. Always as always. And thank you everybody for watching and listening. We will see you next time on AI inside. Take care.