Astra Reactions, Jobs Print, Cybercab Launch | Diet TBPN

0:01John Coogan:Big news in the AI world, obviously. Astra GPT-6 launched yesterday. We went through the blog post briefly. It went up mid-show. A bunch of impressive benchmarks. We talked about Arc AGI. We moved the goalposts because Arc AGI v3 is now saturated, which is particularly crazy because of how detailed those games are when you play them. The thing that stuck out to me, though, was watching the demos of people build 3D scenes in Blender using Astra. Clearly, a bunch of people got access to the model before and were able to fire off one prompt, maybe a few prompts. It feels like it's very, very capable to just fire up Blender and do a full task and model a whole world.

0:53That said, Matt Schumer, who had a couple of the most impressive examples, he has been talking about how the art of prompting is back. Oh, interesting. He had a very specific process where you have like a manager agent and sub agents. And it's still not as simple as just like you can get a good output doing in one shot. but to get some of these really incredible outputs does take a little bit of refinement.

1:23John Coogan:Yeah, and I'm sure a lot of that information will propagate across the internet as it already has. I'd love to pull up some of the images from, or the videos, from the launch from folks who are actually test driving GPT-6 Astra. First, Sharif said, GPT-6 Astra recreated the Palace of Fine Arts in Blender. This is my favorite building in San Francisco because it was built for the World's Fair in 1914. I didn't know that. Where the steam locomotive and telephone were showed off. Wow. It was a time when technology gave us all a deep sense of optimism for the future. I feel like some of that sense has been lost since, certainly.

2:05John Coogan:But I hope models like Astra can help restore it and push us towards a more hopeful future. So if we can play that actual video, you can see. So I imagine that it pulls images from the internet. There's a whole bunch of, obviously, reference diagrams. There's a whole bunch of different ways that you could ground a 3D modeling exercise. You could even pull the blueprints. There might be free models out there. Yeah, there's probably architectural plans. And yeah, it's this odd thing where you could use an AI image generator because the AI image generators are knowledgeable about what the Palace of Fine Arts is.

2:35John Coogan:It could generate a bunch of images there. So there's all these different ways to triangulate it. And then you can also just go on, I think Sketchfab is one of the big free 3D model purveyors online. And so you could potentially go grab an actual 3D model build off of that for your scene. So there's all these different ways to solve a problem. And Astra seems to be good at actually going and doing it. There's a few more that we should look through. There's one comparison, since this is the new Pelican bench apparently. Same prompt, same goal of creating this fly-through video for a real estate scene.

3:10John Coogan:And this feels like something that could be useful. I've already seen on Zillow, there is the ability to restyle images. So if you're looking at a house on Zillow, you can see the interior photos. And then you can say, I'd like to redecorate this in a modern style or in a contemporary style or in a colonial style. And it will go and use AI image generation on the back end to do that. It's only available on a fraction of Zillow listings. It's clearly some sort of feature that they need to turn on. But you could imagine this being something, especially for spec homes that are selling that maybe haven't been, that they don't have the budget to actually go do proper rendering.

3:53Just lowering the cost, you get sort of a Jevons Paradox moment.

3:57John Coogan:There's another one from Thomas. He says, Astra is very good at 3D modeling and the lighting is way better than this one. This is what, this actually looks properly ray traced, properly lit, really thoughtfully designed. Maybe that's in the prompt. Maybe that's just the way this one turned out based on the photos from a Blender scene to then an Unreal Engine 5 walkable experience. So everything becomes translatable. You'll be able to take your tweets, turn them into blog posts, turn them into videos, turn them into 3D games on Steam. GTA-style video games. Basically, basically. And so there's a little bit of a prompt, and Thomas reveals a little bit about his process here.

4:34I would like to see VCs, if you're going to put out an AI blog post, you know, a blog post. Like a market map. an essay. No, if you're going to put out an essay that was clearly written by AI, you should be required to put out a video game that accompanies it. Yeah. That's sort of playable that gets the same concept.

4:50John Coogan:I'm thinking like instead of a market map, make it a Fortnite map. I want to be able to play in the different, you know, town is a popular AI startup, might land on a market map, should land on the virtual Fortnite map. There's some debate over this last one, whether there was like a different harness involved, but same thing, give it a Zillow listing, it can 3D model of the house. So this, it's always with these new releases, it's interesting what makes you feel the AGI or not. Realistically, Frontier Math, Tier 4, I think they scored 99 % on that. Terminal Bench, Science, 0.1. They're just numbers to me.

5:28John Coogan:Like they don't do anything for me emotionally because I don't have any grounding in how hard those particular things are. And also, So I sort of assume that computers are good at math. And you didn't get past algebra and look how successful you are, right? Exactly. So those are rarely. You're like, I've accomplished all of this off of Algebra 1. Yeah. Freshman level math Olympiad. Yeah, exactly. So those are never really fieldy AGI moments for me. But I've actually spent hundreds, if not thousands of hours in 3D modeling software. Mostly in Cinema 4D and Houdini. And it was really fun. I really enjoy making 3D renders and motion graphics.

6:14John Coogan:And it was also useful when I was making YouTube videos. But every step in building a scene requires you got to watch a YouTube tutorial. Then you got to look up how all the different functions work. And then ultimately, you're tweaking settings in literally four dimensions because it's a 3D scene. But then over time, you have to map where the camera goes, where the different objects are moving, how things progress every time. And it can also be very compute intensive. So there's a whole bunch of times where just to preview the scene in low res, you might have to cache and load and pre-bake certain animations so that you can just see what it looks like.

6:52John Coogan:And then you're like, ah, it's not there. I got to go back and change some setting for some sort of smoke simulation or some sort of fracturing that's going on. And so it's fun, but it is extremely cumbersome. And ultimately you're doing this on a two-dimensional screen and it can be very time-consuming and frustrating. And so by the time I stopped spending time on motion graphics, which is around 2021, I'd been building the YouTube channel and I hired a few editors. Some of them were good at After Effects. Some of them learned Blender. Blender was ascendant at that time. So Blender is the open source community-led effort.

7:27John Coogan:Cinema 4D is a closed source piece of software, very expensive, was very expensive, was like over$1 ,000 to get a license that would be good for like a year. Then they eventually went subscription. And so there were pieces of software that were used by Hollywood but were very expensive, very high barrier to entry. So you didn't have as much of a flourishing ecosystem. Blender in 2021 was clearly on this trajectory where it was going to build this ecosystem and become very, very popular. And it did. But by 2021, I was able to hire video editors who were able to go do scenes in Blender. But still, it was a huge ask.

8:02John Coogan:If I was like, for this YouTube video, I was posting every single week, I'd like a motion graphic to illustrate something. And I did this one where we built a 3D battleship board for this U.S.-China military competition competitive analysis. And I mean, it's like it sort of ties up an editor for a week while they're working on this. And this just speeds all of that up. So you're going to get back to like the, is it over for motion designers that worry about job displacement? I think we've been through enough of these cycles that hopefully we're Jeven's paradox pilled at this point. We imagine it will.

8:41You should just pump it to software engineering.

8:42John Coogan:Yeah, exactly. There's a lot more 3D renders. It seems like everything that can be rendered will be rendered. Yep. and more people are going to try to make something themselves and then hit the point where they're like, okay, there's this sort of area where the models can't get it to 100%. Now I need to actually go find somebody who's an expert and work with them. Yeah, totally. And so we're back on the is this AGI question. The Wall Street Journal said, yes, we are entering the era of artificial general intelligence. It's in the journal. They've declared it. Although entering the era, like the door, the house is AGI, the door, and we are entering, we're like almost in the AGI era, according to the Wall Street Journal.

9:32John Coogan:I don't know. Stripe called it AGI. Didn't they call it AGI Jan 1? Something like that. And there's been various calls. Tyler Cowen called it AGI like three years ago. He was ready to rock. I think that was GPT-4. Something like that. No, it might've been 01. I think 01, he was like, this is AGI. But it's clearly artificial. clearly intelligent don't you think he'll end up looking like pretty good in the fullness of time on that call 100 because it clearly is this like you know timeline of spectrum yeah i do like to imagine the the the stripe uh the collisons just like at like just after midnight you know there's champagne maybe some sparklers and fireworks and they just go that's it it's agi it's agi but the question is no longer is it artificial we know it is the question is no longer is it intelligent.

10:17John Coogan:We know it is at a lot of different tasks. The question is, how general is it? Because Blender, as open source, can be downloaded and copied a million times into reinforcement learning environments. We know that OpenAI bought, quote, tens of thousands of Mac minis and Mac studios to do computer use training on. And so the question is, is it only good at Blender? Will it work in Cinema 4D? Will it work in Houdini? Will it work in a small piece of software that does 3D modeling that has fallen by the wayside. Now, there's an interesting dynamic here where I was in the boat of, I could see that Blender was going to be the new hot thing, the best tool that I should have moved, but I already had so many plugins and textures and rendering, and I knew the layout that I downloaded Blender, I opened it up, and I was like, I really don't want to have to get over this learning curve, even though I know in five years, Blender's going to be way better than Cinema 4D.

11:11John Coogan:There's this switching cost. Now, you're in this weird world where it's pretty good if you're a software company where the Frontier Labs bullseye lands on you. And I'm thinking about Slack, like Alex Wang at Meta just moved them to Slack off of Teams where they had like an internal workspace thing. And specifically, he said. Give it up for Benioff. Give it up for Benioff, truly. And specifically, he said Slack is better for work with AI agents. And why is that? Well, it's because the Frontier Labs all use Slack. And so they have Slack as a reinforcement learning environment. And so it got really good at Slack.

11:48John Coogan:And I think that even though there are open source Slacks and you can vibe code a Slack competitor, as we have tested, Slack was the logical tool to pull off the shelf for the AI companies, for the AI agents, for the models. And Blender seems to be that. So the interesting thing is that I do wonder if people will wind up shifting from Cinema 4D to Blender because the models are better at Blender. Or will the models generalize and say, yeah, this looks a little bit different because it's a slightly different piece of software. But I've trained so many hours in Blender that I can also do Cinema 4D or also do Houdini.

12:25John Coogan:Because some of those other softwares are better at certain things. Some of them are better at smoke simulations. Some of them are better at water simulations. There are advantages to each piece. I don't want to degrade any particular piece of 3D software. I love them all equally. There are some reports that it's quite effective at video editing. Cool. I'd be interested to take it for a spin. I wonder that the interesting thing is that most video is still edited in Adobe Premiere. It's a closed source Adobe. They have their guardrails. They might not want to say, oh, yeah, sure, take 10 ,000 licenses, no problem.

13:01John Coogan:pay us$10 a month for that. They might, they may have a different opinion. They may want an AI agent that's internal there. There might be all sorts of different things, but then again, there are probably there. I mean, I know that there are open source video editing suites out there. I think that they lag a little bit, but there are some that are free or, or, or open source at least the other interesting thing. I mean, video editing, like all these things, like taste is so important in these it's, it's, it's often less about the, the, the exact, the exact instructions and more about telling the right story, using the right tools, not creating slop, whether you're doing it yourself or with an AI agent.

13:39John Coogan:But I think that gets to the core question of like, how do you evaluate these model launches? We were discussing the idea that like the benchmark era might be coming to a close in some way, because as soon as the benchmark's created, it's hill climbed very, very quickly. And then it's at 99 or a hundred percent pretty quickly and everyone sort of their eyes start glazing over in benchmark discussions. And it just goes back to like the vibe check. I think the best thing to do when a model comes out is just take it for a spin and mostly point it at something that you know really well, because you don't want to be in the, you want to expose the models.

14:20John Coogan:What's the, what's the Gelman amnesia? That's the one I'm thinking of. You basically want to test if gal man amnesia is triggered uh by you using the model so point it at something that you intimately understand and see if you're impressed and then yeah because you pointed horses what yes yeah yeah we got to do the horse bench again we have a couple couple more ventures let's see comedy i want to i want to see some new jokes shrimp shrimp fried rice bench oh there are other reactions to astro we should go through in the timeline the big one comes from joe weisenthal he says you're never going to believe it but someone one prompted the new model to make a completely new and immersive 3d game uh yes the bar has already been raised the goalposts have been moved uh it's not enough to make a game it has to be funny entertaining and i don't know i i know how i i feel like a lot of people are taking the easy way out with these web-based games playing a game on a macbook pro in a chrome tab that's not real gaming it's got to be It's got to be running on a PC on Steam or on a PS5 or on an Xbox.

15:32John Coogan:It's got to have more to it than just some fun game that you can play around with in the browser. There's also some updates. Token pricing is effectively meaningless now. The benchmarks are meaningless. The token pricing charts are meaningless. Everything's meaningless. No. So Stephen Heidel is sharing the cost efficiency, how you should think about cost efficiency in the modern era. So 3.8 Flash looks 13 times cheaper when measured per token, but Astra is cheaper per task since it's far more efficient. Measure your cost per task, not per token. So interesting. Will we see a decline in – if models get more token efficient, will we see a decline?

16:14John Coogan:I mean we probably won't see an actual decline in token volumes. But, you know, we've seen Sundar Pichai from Google and Alphabet stand on stage and talk about the exponential increases in token volumes. But you might see a kink in that graph or like a deceleration to some extent if models just get way more token efficiency. Like if the next version of Gemini or 3.8 Flash is 100x more token efficient, you might see a little bit less growth in token volumes. but they might be doing more valuable work than ever. So sort of an interesting nuance. You've got to adapt to the modern times, I suppose.

16:54Jack Altman saw the benchmarks and said, Sam, exclamation point, I guess even a blind squirrel can find a nut sometimes. Congrats.

17:04John Coogan:That's good. That's good. We should move on to the jobs report because even though AI is getting better, so is the U.S. employment picture. The U.S. added 162 ,000 jobs in August. Sheesh. That is a lot of jobs. Also, what's interesting is that the Wall Street Journal had an op-ed. Merit of violation. Had an op-ed just, I mean, it just printed today. It went to print online yesterday, but saying, why we'll have to get used to meh job reports. The economy doesn't need to generate all that many jobs to keep the labor market steady, but that won't keep low job growth from feeling dissatisfying. So economists think or they thought that Friday's jobs report will show the U.S.

17:50John Coogan:added seasonally adjusted 53 ,000 jobs in August. The number came back 162 ,000, three times what these economists, what were the economists doing asleep at the wheel? They're on summer vacation maybe. Who knows? That would be a step up from July's unexpected loss of 23 ,000 jobs, which, of course, was now revised upward. So we actually did add jobs. Oh, we got an upwards revision. Yeah, upwards revision. You don't see that much. No. Usually. Narrative violation. Usually. Yeah. That's a narrative violation. There's a whole bunch of interesting dynamics going on in the economy. The Wall Street Journal has been doing a great job highlighting the main street millionaire, the small town entrepreneur, what's happening in the broader real economy, not too focused on what's happening in tech and tech jobs, where obviously a very small employee base can create a really outsized outcome.

18:45John Coogan:Nothing shows that better than hugging face. Couple hundred employees,$12 billion outcome. That's not the nature of the US economy. That is the power law. That is the outlier. But it's easy to track and say, oh, well, like if you're making a$12 billion business with just a couple hundred people, there's not gonna be any more jobs. But that's not the way the actual US economy works. So what is going on with the US economy? The August jobs report is out and it is good. It blew past expectations with the U.S. adding 162 ,000 jobs. Economists polled by the Wall Street Journal said that the U.S. would add 53 ,000.

19:20John Coogan:That comes after both the June and July jobs reports were revised up to 31 ,000 and 21 ,000 respectively. So the unemployment rate remained at 4.1%, which is historically low, which indicates that we generally have a healthy labor market. The jobs added were fairly broad based from the Wall Street Journal. Food services and drinking places were a standout. Let's go. With an increase of 59 ,000 jobs. And people were worried about this post-World Cup slump. You remember this? Yeah. Yeah, and so it seems like whatever happened there, whatever gyrations happened, people are still dining out. People are still going to bars, and so they're hiring.

20:02Well, I think it's quite possible that recent model advances are allowing people to have more power lunches. Yes. It might have historically been a Friday thing.

20:12John Coogan:Yeah. Maybe now they're adding Monday to Wednesday. Sort of a let Astra take the wheel scenario after 1 p.m. in many offices across the country. Could be a thing. Could be that. Similarly, local government education gained 42 ,000 jobs after losing 58 ,000 jobs in the prior months. Manufacturing was among the sectors registering job gains, as was health care, as has always been the case. Health care continues to boom. The information and finance sectors shed jobs, while economists are divided over how artificial intelligence will divide the market. Those sectors are generally viewed as among the most exposed to technology.

20:49John Coogan:At a minimum, the report is an argument against the Fed lowering rates in September. Before the report, markets have been roughly divided on the prospect of a hike. It also seems that the AI jobs apocalypse is still on hold, at least for another month. And of course, we'll discuss this on the show. The great Dyson debate. T.J. Parker kicked it off. He said, apparently unpopular opinion. Dyson products are generally not very good. Strongly agree, though my girls love their air wraps, says Bryce Roberts. A couple other people saying, yep, my wife got a vacuum. It was terrible. Felt like a total psyop.

21:28John Coogan:A lot of people are upset. The Matic founder's chiming in. I got an alternative for you. Oh, he's saying that it's likely ruining your rug slash carpet. Oh, interesting. Dyson rug shed and Dyson suction is very adept at making it shed faster than necessary. Okay. Okay, I'll start with what I think the Dyson business is incredible. The story is incredible. I think that the making things that look like halo weapons for stuff around the home is a very cool way to differentiate. And I appreciate the overengineering, like everyday objects. I'm not saying I'm buying into it, but I appreciate that someone's doing it.

22:12Somebody needs to make a$500 toothbrush. Okay. But that being said, I saw this and I was like, you know what? That does resonate because every time I have a Dyson vacuum at home and I've had various Dyson products over the year, vacuum in particular, it just feels like a little bit like it's like creaky. It feels a little bit cheap. Yeah. And I always want to put it away because I don't like this, like halo weapon, you know, a home cleaning device doesn't exactly fit into the overall design.

22:45John Coogan:Just say you don't live at the Citadel. Just say your house doesn't look like it would be welcoming to the flood. Come on. You got, you got to design your whole house to fit the aesthetic. Yeah. You got to have a Banshee out front. A Banshee out front would fix me. I think you forget how bad it was before Dyson. It was so bad. Vacuums were so rattly and clanky. And they used to have bags on the back that you would have to take out and they'd have this felt inside and the stuff would be going everywhere. Like this is such a step forward. Maybe you could say that like it's plasticky and it doesn't feel like super rock solidly built now.

23:26John Coogan:But it moved the whole category so far forward and so far into the future. Just the fact that you don't have to plug it in. It has a battery. That was an innovation. Like most vacuums, you would have to take it out and then you would unwind this long cable. And I think they still make some of these, but you would have to find the nearest plug, plug the vacuum in, and then you could only vacuum as far as the cable would go. And then you'd want to move to another room. So one of my early chores, I think, I don't know why I have so many remedies for this, but I guess I vacuumed. I have formative memories of like trying to get to a spot.

24:06John Coogan:And you're like, oh, and you're trying to extend it. And then you have to like get the cord over something else. You're trying to be as efficient as possible. Yeah, of course. You don't want to move the cord in two seconds. There's four plugs in the house that I need to hit to vacuum the whole house. And so you've got to optimally – it's a traveling sales problem all of a sudden. It's NPR. You've got to plug in over here, reach under there, get under the couch, go over there. Then you've got to move the cord over the couch so you can get to the other side. Then you've got to move the cable all the way to the other side of the room so you can get to the other side of the room.