GROK 4.20 and the "SOCIETY OF MINDS"

0:002% of the world can code, I think, and that's about to flip to where 100% of the people can create software on the map. I don't like the idea of a huge war machine to actually be intelligently moving in a direction that we have to then counter with even more intelligent, non-war machines constantly. With agents like this, anybody can build fairly complicated pieces of software. Just need to work as an entire planet because on the other end is everything we want. When you just allow the entire more people, where that's the entire world, there's more people to contribute their ideas, stuff just blows up and, like, you find genius in the most unlikely places. What up? What's going on? Welcome back to another episode of the Western Dylan podcast. We got some big news for you today. It's in foremost, GROC 4.20, then we also are talking about OpenEI and OpenClaw and tons

1:00of other stuff. I actually have something cooking that I wanted to talk about, but I guess Dylan, do you want to start us up? Yes. I'll start with the updates. Yeah. Yeah. So pretty fascinating. Not everybody seems to have it all rolled out right now, but there is a new GROC on the market and it has swarm intelligence. I guess it's got four different agents, swarms that are in. What else do they have for the headline here? Specialized agents work together to solve complex problems. I don't know. Let's go. Maybe I'll go experiment with it. Let's see what else. But yeah, GROC 4.2 is the newest one and it is out now. I'm looking for a new model. I'm, you know, the ads are starting to roll out on chat GPT, so I think I'm kind of at my last, my last month of the $200 plan, I'm going to be switching over. So I'm excited to see if GROC 4.2 actually lives up to the main reason I use it for kind of like tracking down news. It seems like it's plugged into Twitter.

2:01It should be really good at that and how intelligent it can be when it comes to summarizing articles and like helping surface the stuff I care about the more kind of below the news news that I like to cover on my channel. So I'm excited to play with it. Absolutely. Same here. So it's went live shortly before we did shortly before we started this episode. So we're going to be kind of trying to catch up in real time a little bit. So apologies for that. I'm cracking up. So first of all, to all the people that leave comments, all the people that leave on this channel, on my channel, I love you all. You guys are hilarious. So every time I mess up, wow, did I get dragged through the mud? Livestreams are just we had some guests on the live stream. We interviewed the two ex-Google guys from the SVAC podcast and the whole time like chat please. Just don't be an animal. Just don't like come on. Keep it together, people. Keep it together.

3:02Keep it together. Keep it together. They're going to get so worried. We need that professional here. In the beginning, I was like, okay, chat please be respectful to the guests. I know we're all relaxed during our regular podcast. This is different. And some people in chat were kind of surprised. Like, dude, who do you think we are? Some of you are already created here. To those that are not having seen my last video, I'm wearing some sunglasses because as I explained in my last video, I have some scratches on my face and it doesn't look great. And how I explained what happened in my last video, I said the reason I have scratches on my face is because I landed on my face and I kind of just left it off at that. So of course, tons of comments. Let's see. Wes and Dylan had a nasty breakup. That got a lot of... That was like the very first video I made on Sora was us breaking up and fighting. So yeah, maybe it's like one of those things where whatever happens on Sora then becomes

4:04reality, you know? Oh boy. A prediction for the future. Oh god. Yeah, what happens in Sora is like a pre-simulation or whatever. I have the... I have the... I have the... It feels like a lot of science fiction is coming to life, you know what I mean? A lot of these things I never thought were going to be real, seem like they're around the corner. So maybe... One person said, weak twice, if AI hurt you, I can't wink. That's the problem. Yeah, I will hurt you again. I can see everything. The CIA handlers took control of him. Sam Altman did it. This is a neural implant scar on your forehead. That actually, now that I think about it, yeah, it kind of does look like they've been planted something in my head. Anyways, so... I hope you're okay. You should have downloaded the not falling on your face, open glass killer. Dude, I got to rate someone. Dude, what did you have? Just like molt book, two in both hands, laptops everywhere. My pain is... Okay. That was extremely funny.

5:04So yeah, anyways, let's rock and roll here. Let's open up with 4.2. Oh, let's open up with some AI news. And then maybe in the end of this episode, we'll get to some of the weird stuff, because I'm sure not everybody wants to hear the personal stuff. But we'll get to it in the end. So really fast, let me just kind of hit the main highlights about the GROC for 20 release. So this is the beta. It's starting this week. Just a few hours before we started recording this, started hearing rumors on Twitter about it on X. And it's seemingly very interesting so far. Now of course, we've been hearing a lot of news, a lot of stuff coming out about it beforehand, right? So in the alpha arena, it outperformed all the competitors, yielding a 12% profit while I think every other model was negative. And that was very impressive, like we, I think we talked about in one of these episodes.

6:06Now it doesn't prove that it's a superhuman equities, a trader, investor, et cetera, because over time, it could just kind of revert back to the mean and you would have to run it for months or years before you knew for certain. But it did look kind of interesting. So it was kind of like one data point. And later, Grock for 20 also was really good at predicting one of those, like the prediction markets where like who's going to win the Super Bowl, who's going to win this, like kind of like those events, it was better than average, better than other models. Also shortly before this few weeks ago, somebody out of a professor researcher out of UCI said that Grock for 20 and early access was able to create a new curve for the Bellman curve theorem. I'm not sure exactly what the terminology is, but basically seemingly it's able to do some pretty complex automation of the theorem discovery.

7:12So I did a whole video about that. So I don't want to go to many details, but it seemed impressive. It seemed like there's definitely something there. And there's one more test that it really ace that I'm blanking on. But again, this was all before kind of like the main release. And today is when it's rolling out the beta version, it has a multi agent swarm. So looks like there's, you know, four agents collaborating together, talking together, discussing everything before giving the answer, which interestingly, I also used that with my open cloud agents, I call it society of minds, basically tying in Gemini and Claude and, you know, the latest chat GPT model, et cetera, and Grock, and howling them all talk together to come up with solutions before presenting those solutions. And sometimes it works like insanely well, maybe we'll talk about it later. So definitely, while this is not a new idea, this idea has been shown to work extremely

8:16well in the past through research, et cetera, and anecdotal kind of evidence. It was built on the Colossus and the Colossus 200,000 GPU cluster. It's the fastest processing speed, deep integration of real X data. Apparently it's multimodal, it has multimodal improvements, dynamic in video insertion, slow integration, rolling out to Tesla vehicles across UK, Europe, US. So that might be interesting. It's going to have personalities like storyteller or unhinged. The unhinged mode is unhinged. I was, I don't know if you've ever tested it, I tested it on a live stream. The first time it came out, I had to shut it down. I'm like, nope, because I don't know what you're going to say next. I can. Wow. It was, yeah, I got a little bit kind of worried about how it's going to affect my account. So there's subscription subscription tiers, super groc, premium plus, and looks like API

9:20access is going to be available, apparently recently set records for processing trillions of tokens per week. Okay. So that's kind of like what we know so far. Yeah, Dylan, I mean, what's your, what are your thoughts about this thing? Well, yours, would you call this society of minds or would you call it, would you bring all the models together? Yeah. So it was originally, I think, a paper in psychology back decades ago about how our brains are actually a society of minds. It's not one thing that's up here. It's multiple that kind of come together to form whatever. Then it was a Google paper, and then I stole that name for my chatbot. So yeah. Okay. Yeah. Now, I mean, that's a fascinating idea having your multiple agents sort of talking of it. You know, I assume if you're thinking about something inside of a groc, they, what do you know about the different agents? I'm guessing one sort of is trained to be more logical and one to be more like higher level thinker and one to be more like some kind of news aggregator and they sort of talk

10:21at different levels. But when you, what you're talking about is the full models talking to each other and they're trained, you know, from different labs with different kind of data sources and different philosophies. So it'll be interesting to see they're not necessarily meant to talk together to solve a problem, but they're smart enough that they do. So what were some of the examples? Yeah. I mean, this is such a fascinating thing and it's a little bit off topic, but I think it's a very fascinating kind of rabbit hole to go down into. So let me give people kind of maybe a high level overview. So there has been tons of research papers kind of looking over the subject and attesting how these models perform when you just kind of put multiple agents together and allow them to collaborate, collaborate. Now, what's interesting is before kind of in the early days of these models, this was almost like built in that whole thing where I forget now what they were calling it, but they would do, I guess, sampling, right? They would sample model a hundred times.

11:22And let's say 80% of the time it answered had one answer and then there was a sort of distribution of answers for the rest 20% of the time. Most likely, the most common answer was the most likely, right, the most likely to be true, which I kind of think of it as like imagine you're in a city, even there have been two before, you're asking for directions, right, you ask 10 different people and eight of them tell you, you know, you go to blocks this way, take a left. And then the ninth person is like, oh, it gives you some crazy directions and the tenth person says, oh, you just jump off the bridge, right? Which one of those answers is correct? Probably like the eight people that told you the same answer, that's probably the correct and right, they just make sense. That's most likely the correct answer. Where's the other crowds? Yeah, that's kind of exactly what what what what how it works here, since it's kind of like probabilistic, the cast, the answers, how are you when I think of it? If it's coming up with the most common answer that test might be the most likely to be

12:25the correct one. So that's that's how a lot of this started, but later when these models got good enough, what researchers were doing was putting them together to chat, to talk, to work towards some specific outcomes. Chad Dev was one of them. That was the first time when I was truly impressed with how well they were able to create software and the troubleshoot bugs. I was like kind of blown away because that's the first kind of taste that we saw of that. And two, there was a great paper where they took a bunch of them, they made them play Minecraft, so they made four of them. And so the game tasks and then they talked in game and then played Minecraft. And so it was very interesting because when one would get kind of distracted and say, oh, I'm going to run off to the other three were like, no, hey, Bob, Bob, like focus, we're chopping wood because we need to build the that is Bob's like, oh, that's right. So is this kind of like, they would communicate, they would bring distracted people back on track. And basically the sum of the parts were, was the expression the, the sum of the parts

13:30were created in the sum, yeah, or whatever, yeah, basically the four of them created better results than just adding each individual piece or any given agent. So which is interesting to think about because that's not like a, you know, it's obviously we're dealing with something different here with these neural nets and stuff like that. When by working together, they can create something better. One fascinating thing about that paper is towards the end, they've really buried it deep in there because they said at some point these things, they needed some leather or something in the world of, of, of, of Minecraft. And so they end up just like rating a village and like, you know, just burning everything down and getting what they needed, I thought that was hilarious, but that's either here or there. And you have a big enough group, I guess, huh? Yeah. Yeah. It's a point they just, I guess, one of them convinced that everybody, yeah, the pack hunter is, I guess, when you train them that way, like, when you to do this guys, like, otherwise,

14:30how can we get leather? In my sort of experience, so I had the OpenClaw agent itself create this society of minds, get everybody in, in there, and I had the API keys for, you know, the four big models and they all chat it together. As an example of what that looks like is, I asked, um, so clawed to create a little tracker for YouTube videos. I wanted to see when YouTube videos came out and just kind of see how well they're doing get a little summary. And it did. And it did it very well. It was very functional, but it burned through a lot of the YouTube kind of API credits. So it was very, I looked at it like, man, that might be kind of expensive to, to keep track of it, because you keep having to do these calls to check. And if you want to have it kind of in real time, then you got to do like every five minutes or every 10 minutes, I was like, man, this is going to be a little bit expensive. But then Gemini came in, right, because everybody, all the models were kind of looking at this problem that I gave it and kind of suggesting things and looking at the code and giving

15:33suggestions. And Gemini comes in and it's like, like, guys, why are you using API calls on this? Just ping their RSS feed. It's free. You can do it infinitely often. So just use that to see if there's a new video. And if there is a new video, then use the API key to check for views. And I was like, oh, my God, like, I was like, that's, that's brilliant. And now if you're a developer that works in those environments, you kind of know these tricks, right? This isn't like rocket science, but it's, but still like, it's not knowledge. I would have had without having some, you know what I mean? Some experience working with those. But so the key here is that cloud created brilliant code that worked and satisfied with like requirements. But because we added multiple models, Gemini had knowledge that just made it like super efficient and basically free or as close to free as possible.

16:36Like it went from like 10, 20 bucks a day to pennies. And so that's an example of where it's like, okay, you can see how well that works. Yeah. Well, I mean, there's so many times it seems like in my life when I'm solving a problem and thinking over it a couple times helps. But then sometimes leaving and coming back, I just have a different mindset. Somehow I'm thinking more broadly or I've kind of, my head is in a different context. Yeah, it went to bed and I woke up and I'm fresher. And you can just see an answer that you were just never going to get to by pounding the pavement like on the idea originally. So I could see how someone like Gemini might come in and be like, anthropic like, you know, this is amazing what you've done, but it's all been inside this box. Why don't you just solve it this way, you know? Yeah. And I mean, if you think about it, it's, we kind of see that live happening with, I mean, open-closed to thing that's in everybody's mind. But any open-source project, right? It's like, why couldn't anthropic or crock or Google or anybody else build something

17:41like open-cloth? I mean, part of it was, I think it was like really sort of like safety off, you know, full auto mode, whatever. But you know, it's obviously that they wanted to or take, for example, Manus, Manus AI. So Meta bought them for, I forget how much, but it was like a big chunk of money. Why couldn't Meta with all their resources build something like Manus AI, right? You know, they have brilliant people on staff, et cetera, and tons of, I mean, you know, opening eye coming out and beating Google at their own game, basically early on. Again, there's a million examples like this where it's like, when you just allow the entire more people, whether that's the entire world or just more people to contribute their ideas, stuff just blows up and like you find genius in the most unlikely places, which is why. And I think this is very much kind of underappreciated with agents like this. Anybody can build fairly complicated pieces of software.

18:41Maybe if you have time, I'll show you what I've been building without writing a single line of code. And that is, people don't realize this yet, but it's available to everybody in the world now. We're going to see some wild stuff because it used to be like what 2% of the world can code, I think. And that's about to flip to where 100% of the people can create software on them at. Yeah. How different. And automate workflows and build systems in their own life to bring in information that make decisions like it'll be absolutely fascinating. But I do, I do think we should actually look at the stuff that you've got to show. Before we jump to it, Peter Steinberger, who was the founder of OpenClaw, and he put this kind of open source framework out there and then you plug the model into it and we've been through everything from Malt Book to Malt Hub to everything else you can imagine has been happening.

19:43And now you've installed this on your own computer. When you think about what he's built and the kind of person he is from his projects that you've worked with. What do you think he's going to be like inside of OpenAI because he's now working with Sam Altman and he now has access to a lot of different tools. Does it change the way you think OpenAI is going to kind of evolve over the next couple of years? You know, it's an interesting question because I do feel like OpenAI gets sometimes an unfair kind of representation. I think that they believe a lot of the stuff that they're saying. And in the industry, people that work with them do perceive them as Open and trying to do their best. And in fact, recently I feel like the sentiment with Anthropic kind of like really shifted right? Because Anthropic is the one that's shutting down these projects, the one that's using a lot of lawyers to push people around instead of working with somebody like Peter Steinberger

20:47I don't know why I forget his letters. Steinberger. Yeah, Peter with Peter, the creator of OpenClaw, I mean, he's just such an obvious kind of now quickly became like a staple of the AI culture world, right? And you know, if you're a meat like an AI lab and you just like try to crush whatever he's doing, wow, does that make you not seem like the good guy? But at the same time, just today, you know, we're seeing Anthropic is, I guess they're fighting with the Department of Defense or the Department of War or whatever the Pentagon over how Claude, yeah, how Claude is going to be used. I haven't looked at that yet, but that's kind of trending as of this morning. But I think Peter's going to do exceptionally well at OpenAI. I think that money wasn't an issue for him, right? He sold the previous company for 150 million, right? So he's not, he's not starving. He's not probably not like motivated to just get a little bit more.

21:53I think he enjoys working. I think he enjoys working the space. He's fascinated by it. And I think opening, I had everything, you know, they agreed to keep the OpenClaw project alive as open source. I think they're seeing the same vision that he has in terms of like where this agentic environment is going. They have those cerebrus chips, which I think he mentioned as being one of the sort of like things that might be interesting because just the speed and just amount of compute that he's going to have access to. And of course, he's going to be able to play with the latest toys, the latest model releases, et cetera. He says that the best coding agent is kind of like the codex series of agents. And so yeah, I think this is just like the perfect partnership. I'm sure Mark Zuckerberg gave him a huge offer. I think there's a number of other people that gave him offers and he went with the one where where he felt he was going to build the best thing possible. Yeah, they get to that.

22:54Yeah, there's definitely something to be said about speed too. I know there's a codex series now that's ultra fast and that does change the development cycle. I did want to throw one thing in though just because of I ended up watching like Peter's interview with Lex Freeman. So I kind of got the whole deep dive of the anthropic name change. And like just for the record, like when when anthropic reached out to him, they didn't do it with lawyers. They did it just a person at the company did it and so you know, like in most cases when people do like a cease and desist kind of things through their lawyer, they do it because they want to make money or they want to own an IP and like not lose out on the future revenue. Andthropic did it because they were concerned that installing both book and people thinking it's an official cloud product would would be safer when it was clearly an experimental open source project. So I think their point of view was more that they were concerned about like some sort of enterprise and regular people thinking it's like a tool that has like robust guardrails

23:59on it and it not being there. But interesting, that's a good point. But I wasn't aware of that this what I've heard was I mean, it makes sense. But yeah, so that's that I think that again, yeah, maybe then again, too, they're like, hey, we own cloud, that's our name, you know, so maybe there's been an aspect and I think lawyers would definitely have gotten involved if he if he didn't answer. But it wasn't a lawyer originally who reached out to him. Yeah, I think it's also those things that the internet we sort of the royal we as a whole, we tend to overreact for things that we care about. And even a small sort of perceived misstep gets like blown out of proportion. I mean, we want to see open source projects. We want to see things like open claw, right? For us to keep our data for us to not be beholden to these AI labs in the future, you can imagine pretty scary situations. And so when any one sort of pushes against that or is being perceived as an aggressor

25:00against that, in his case, that's what Anthropic was, it was perceived to be attacking this incredible open source project, right? However, they went about it where they did it correctly or incorrectly or what their motivations were. They had a large sod as, okay, we need to really like, you know, perform the mob and yell at them for this so that in the future people, you know, kind of disincentivized, I got a behavior. So. Yep. I agree. Yeah. And that idea about the codex series. I should get over the name of it. I should look this up. But I think one of the new codex that's coming out of open AI for coding has an extremely fast token. It might be on the cerebrus chip or something, but I know that it's coming out. It's not a super deep thinker, but it's a fast coding model. And you know, the idea of watching the code instantly sort of update gives this gives you much more of a real time feeling like if you ever use one of the image generators that works really quickly, like they're half second generations, usually not that high

26:04quality. But you know, when you're prompting, sometimes you're looking at an image before the prompts even finished, you know, so to imagine the same thing with code, you know, coming soon is a pretty wild idea too. And I mean, we've seen Google release. I don't know if you saw it. It wasn't that big of a deal, but they released the diffusion coding model. Oh, yeah. Yeah. That's a diffusion is so much faster, right? Yeah. The mix token. Yeah. I was it was it was weird. Yeah. Because it was like just boom. And the code appears. It's not as good as the lambs, but that's maybe just because there wasn't as much sort of research and money thrown at it. So I quickly opened up GROC for 20 beta, I got to think about kind of like how to really test it because I don't know if it's necessarily a great coding model or at least, because usually when when the new iteration of GROC releases, what are the pointers? I don't know if it's going to be competitive with Opus and Codex, the top of the line models

27:06right after that. And I think Elon Musk you encountered for that. Yeah. But it's better than drop again some ways, but they have something special with coding he said. Yeah. So we have to kind of think about how to test it. One of the things that's great at is getting real time news. I actually asked about the anthropic and I think it's the Pentagon disagreement today. And so very quickly it's thought for 29 seconds. And it gave me what seems to be a pretty good outline. So you know what, this might be actually pretty good, not I think about it, wow. So it has a 228 sources, which is I've never seen anything like that in 30 seconds. It pulled 228 sources and just output the answer. I am so far I'm impressed, but do you have any ideas for how to test it? You can use some of those classic ones like you could ask it like explain blockchain in exactly 10 words, and don't use the letter E.

28:06Oh, I have, yeah, those are yeah, let me let me let me kind of think about that. I'll throw it a couple a couple tests here while while we're talking. But let me quickly sort of maybe let's cover the anthropic and Pentagon disagreement. And actually I'm curious, just kind of maybe just give me give us a quick outline of how you feel about autonomous weapons, like for example, AI and autonomous drones or anything where AI is controlling something that could end human lives, et cetera, just fundamentally do you have a strong feeling about it? And yeah, so yeah, I was like ironically, I was going to be like, oh, yeah, I'm all for it. Autonomously killing people. But no, of course, like obviously the opposite, I mean, the, yeah, I mean, that's such a such a deep question. I mean, I guess looking at what's happening right now is anthropic has a very advanced

29:09model and they do make money from working with the military as do all the other big companies right now and slowly almost all of them Google open AI, they have been removing wording consistently over the last six to 12 months that sort of says, I'm not going to do this or that. Like the goal posts are absolutely being moved and these tools are being used inside of the Palantir and in military settings. And it seems like a anthropic was, yeah, a tropics cloud was actually used in the Maduro capture and that kind of came to light and, you know, they said, Dario Amade said, like, you know, no more, like I'm not going to, you can't use our models for a decision where a weapon is, is fired without a human in the loop. And then there was one other thing about how it was implemented. But it wasn't like a complete cut off, but it was like, there's a couple of lines here that like we, we can't let you go beyond.

30:10And the Pentagon said, basically, we'll punish you for that. Like then we're going to drop this 200, I don't know, whatever, it was 200 million or to $2,000,000,000 project or whatever they had. And I can't think of a better good advertisement for anthropic. I mean, to me, that was like the most beautiful reason to like support their product that I've ever heard. So I was like the exact opposite of kind of what we were talking about before, like it showed that they actually were willing to stand some ground and spend some money for something that wasn't totally unreasonable. Like we're just asking for like the weapon not to be fired autonomously, you know, but I don't know. I also don't want a weak military. I don't want to pretend the world doesn't work the way it does and I don't want to fall behind on a world setting either. So it's a very kind of nuanced answer depending on where you are with with it being used. But I worry about privacy, I worry about weapons. So let me, let me try and experiment here. So let's start with just on a scale of what you tell how strongly you feel about these

31:13autonomous weapons and if any sort of new ideas or details or anything else, if that changes your mind in any way, shape or form. So on a scale of 1 to 10, 10 being strongly against one, like whatever, like how strongly are you against, for example, just autonomous weapons, like drones or tanks or whatever being unleashed to go there and hunt down human beings, like 10 being massively just against that. Yeah, my pretty much basically against that. I think if we have this kind of intelligence on the planet, there's got to be a better way to solve things than just killing someone. You know what I mean? Like, if you can, if you can hunt someone down and shoot them, can't you just hunt them down and handcuff them? Like, can't some sort of drone with handcuffs, like get somebody, you know, I don't know. So I'm pretty against it. So let's say, we'll 10 out 10 or like 9 out 10, I mean, I guess 9 out 10, I mean, I'm sure there's some situation where somebody told me and that's what I'm curious about. That decision go, but yeah, 9 out 10.

32:14And that's what I'm curious about. Is there anything that makes it more okay or less okay? So here's the question. So oftentimes, like the counterpoint that they give is if a drone is connected to some remote operator, that means there's a link and that link can be hacked, it can be disabled. So if you have some weapon that's communicating, but is able to do it, like what if it gets hacked by using that as a, you know, an attack surface, so to speak. And then it can be turned against the civilians or against the good guys, whatever. I mean, this is obviously sort of whatever we're kind of just keeping it broad strokes. Does and so being it fully autonomous, it's more secure. Does that change your opinion anyway? Does it shift it one way or another by a big margin? Is it being totally, yeah, yeah, I mean, like there's models that if they are running totally autonomous and they have this incredibly high confidence that taking some kind of action

33:18to hurt somebody is going to save other people, it's kind of the trolley problem. Yeah, I don't think I could just say like never, ever, ever feel like I've had to go and just change my camera on me. But yeah, there's just like such a rare, there's just such a rare situation where I think that would come up. You know, I can't, I can't say I'm totally against against it in all situations. Yes, yeah, for me, I was like always very nervous. I was scared about like scared where it could leave. And I understood kind of like the counterpoints about security and safety. But I was like, man, still seems like just a slippery slope. The only counterpoint that I've heard that I was like, huh, kind of made me think about a little bit more was surprisingly, was Arnold, Lucky Palmer. I don't think Arnold's in there, Arnold Palmer, is that Lucky Palmer? Not the golfer, not the drink, the, I think Palmer Lucky or Lucky Palmer?

34:20Lucky Palmer, yeah. So he was, he was one of the developers of VR early on, pivoted to now more defensive tech for the, he's been an autonomous fighter now. So one of the things, and of course, controversial character, very outspoken, you know, so let's forget, because oftentimes I feel like people dismiss ideas based on whether they like the person or not that's saying it. Forget that for a second. One thing that he kind of said was like, look, we've had autonomous weapons for centuries. She was like, what are you, what are you to centuries? What? And he was like, yeah, I mean, if you think about something like a landmine, right? So you, you bury it and then it sits there until somebody steps on it, right? And it blows up. That's, I guess, yeah, I guess it's an autonomous weapon. And obviously there's no sort of human in the loop who wants it's armed. Like a human doesn't, you know, get to check it. Like, do we blow that person up or not? No, it's like, if there's pressure, it blows up. And even if you go to medieval times, there's examples of various traps

35:20during sieges, whatever they, I guess you can call autonomous weapons. So he's saying it's not the autonomous part that's new. It's the, the intelligent part that's new. And of course, I mean, some people, I'm sure people are aware that there have been places in the world where I think Bosnia was like, that took kind of, I think really, really bad for this, where during the war, they just put a lot of landmines in the, in the forest, on the road, et cetera, et cetera. And, you know, like decade, decades after the wars over, whatever conflict that was settled, you know, there's still families and kids that are just walking. You know what I mean? Walking to the forest or whatever, get blown up. And so when you, when you look at it from that perspective, it's like, yeah, I mean, in terms of, if you're strictly talking about autonomous, that is, I think, literally an autonomous weapon, right? There's no human loop, nothing like that. There's no checks. It's just, it's just there. And it just takes out anybody under that condition. With these smart systems, potentially, we can see it could maybe reduce civilian casualties,

36:30because it's, it's smarter, right? It's more targeted, et cetera, et cetera, et cetera. And I hate that argument because it's, it's such a good one. It's such a good one because it's like, wow, like, how do you in begin to make that decision? Because like, what if there's, like, just unequivocal data that this helps prevent unneeded casualties, like what you're saying? What if it like puts the handcuffs on them when that's appropriate or shoots them when it's appropriate? And we see that there's less civilian damage, there's less building damage, there's less suffering, right? So it doesn't wound somebody, you know, if they need to be eliminated, they get eliminated painlessly, right? I mean, you could see how somebody could present to you some arguments where you're like, wow, this thing is more humane. And then you might be okay with autonomous weapons. I wish point, you know, smart autonomous weapons, at which point, of course, right? We have

37:33smart autonomous weapons. And then what happens then is kind of the question. So I don't know, like, yeah, a couple of thoughts. So one is, one problem is that the risk kind of moves up the chain. So let's say we do have somebody from a political point of view who says, like, you know, this group needs to be targeted or like, you know, there's really a database of everybody who's an illegal immigrant. And it's just like go put handcuffs on all of them right now. And the drones just go out and do that. Now you're trusting that the database is correct. That the spying system is correct. And it's still a different problem. You know, it's just kind of like, now it's a little harder because it's up the chain. And everything is happening like automatically on the way down instead of looking at the individuals who are making mistakes and doing the right thing. But, you know, US before, if I like, if I want the weapons or how much against it, I am, but there's also the question of like, do I see a world where it doesn't happen? And the answer is definitely not because in game theory, you are in a situation where if one side in a conflict says, hey, I have already

38:41deployed, you know, five, one million autonomous drones. Each one, you know, has an explosive on it. And it will, you know, and it will act autonomously. Like you have to respond with the same thing to deter them from deploying it. So there isn't really going to be a world where I feel like this isn't going to happen. You know, so I still prepare for it. Even if it's not something I'm in favor of, it doesn't necessarily mean I think that I'm going to be able to stop it or we should all think that. Yeah, one great sort of way of looking at this that I think really helps people maybe kind of wrap their minds around it is like when we think about intelligence, it's hard to understand what that means as it keeps going up and up and up. Like we start like once somebody gets way past your intelligence level, you start you stop being able to understand how brilliant they are, so to speak. Yeah, I just had the situation where I was there was this, what was

39:42it called? I think it was called C launch where this it was this burnt out oil tanker and they used it to launch rockets from at some point. I was very fortunate. I got a tour of it way back back back in the days. I was like a kid basically. And it was interesting because it was like this room of brilliant engineers and rocket scientists or whatever, right? And then to me, they were all brilliant, right? But to them, there's this like one person that was like, oh, he's like the really like mega. So all these people much smarter than me looked up to this one person that was like, oh, he's so smart. I could not distinguish, right? Because to me, they're all kind of like, do you know what I mean? Absolutely. Yeah. I'm like another difference. Yeah. Between it's like super intelligences. Like, like I trust you in ones 20 Einstein's like, what do I know? But so like when we're talking about this AI getting smarter on the battlefield, people might not quite understand what that means, right? Because it's kind of hard to imagine. Like, what does that mean? But there's another sort of way of looking at it that I think is a lot

40:46more intuitive. Like, so imagine you have two players. One is okay at chess, right? decent. And one is just an absolute grandmaster, right? And so you can imagine them playing chess. The grandmaster probably beats the other person. And if this person keeps getting better, maybe they'll win some percent of the time, right? And if this grandmaster at some point, like, if he's just beating the novice every time that he gets twice as good, three times as good, maybe, you know, he's still already winning every time. So there's not much that much of increase. Now, imagine that the novice gets four moves for every one move that the grandmaster makes. Who's going to win? Right? It's a novice in chess. It's it doesn't matter how smart you are, how brilliant you are. If the other person makes four moves, they will and you get to make one. Like, they're decimating you before you have a chance to react. They're executing entire

41:48like strategic attacks while you're just setting up the first piece. Like, your intelligence is canceled out by the sheer speed of the execution. And what's the equivalent in geopolitics? Well, in war, speed, I mean, speed, speed, speed. Specifically, like, well, later, I do want to kind of mention the thing that I'm building. And basically what it does is it takes news and it compiles it. Everything's built by AI agents. Everything's maintained by AI agents. And one thing that I've been kind of running into, I'm like, how do I create a system where it intelligently ranks, how important news are, and invariably part of my mind goes, oh, well, I can just jump in there and check. Like, it sort of like runs it past me. I'm like, this was good. This is bad. Right. That's kind of my first go to, right? Because I'm like, I'm better at making those decisions, right? Right. You're like, let's use WordPress. And it's like, no, right. But the thing that I

42:48keep running is like, no, because I have to sleep. Sometimes I'm not going to be able to do it. It can make those decisions in a split second, a split second after it happens. So if I'm able for it to make, you know, a move that's, it doesn't have to be as good as I am. It has to be just good. You know what I mean? Or not even that I want to move for everyone that you do. It can still, right? It can still win. It can still, the thing in the aggregate can be a lot better than me with my limited, you know, ability to make decisions. There's only so many decisions each of us can make at a day. This thing can do 24 hours a day. So it's like it's one of those things where just speed and the fact that it's automation will beat out a human any day of the week. There's this other thing that they talk about, which is like cognitive offloading where they do for like for jet pilots. Stuff moves so fast that they can't possibly be expected to think through every single thing while they're flying, you know, a freaking jet plane. Right. So there's

43:51this idea of cognitive offloading. So offload as much of the thinking to the machine as possible. And you're only kind of like that top layer brain power that's required when needed for slow decisions. Oh, we didn't talk about my credit card being lost shoot. Anyways, to all the people that have helped me with the credit card issue, thank you so much. I was incredibly surprised. Somebody called the credit card company to apparently cancel it. Absolutely. You people are the champions. Thank you so much. Absolutely blown away by how honest and indecent. I mean like 99% of the people are right. There's just that 1% that was like, Oh, I'm going to go shopping, shopping spree. Right. But the rest of the people are just absolutely incredible. So no, but no money was lost. Everything was fine. And I guess we'll talk about it some of the time. But yeah, it's I assumed because it was just like, what is it? The Occamstrator? What Occamst Racer? It was the obvious thing. Yeah, you gave it to some experimental AI chatbot. Yes. And there were

44:57stories about information getting leaked. How do you think your credit card got leaked? Obviously. But if you watched that video that we did last time, like I was kind of like, I was, I doubted it that it was that. Remember I was like, I don't know because what about my API keys? But I was like, it's just the most obvious explanation. But no, actually what happened was it was a human. It was a human stupidity. That's me. You know, leaving that credit card exposed, and it was human greed or whatever you want to call it that. Okay. So wait, can you explain when your credit card got exposed? Because I was on that stream too. And somehow I don't know, like, did you hold it up? No, no, I held it up. I can hold it up right now. No, no, no, no, I held it up. And it was one of those that on the front has nothing, you know, just the logo, everything's in the back. So you were safe. Well, no, that's that's safe. Like what are they going to do? You can't do anything with that. Just just a piece of plastic. And then on the stream, I typed it

45:57into into to open claw, right? And then it tried to check out and eventually did it figured everything out later when I was demonstrating everything that I've built. I'm going through the chat logs. And I'm showing everything. And so the chat log is like massive. And I was very careful. You know what I mean? I blocked out the stuff that I didn't want there. I tried not to show the stuff that I wasn't there that I didn't want there, et cetera, et cetera. There was like, I think like a frame, one frame, right? So whatever it is, it's like 60 frames a second or 24 frames a second, right? Twenty four, whatever it is. Yeah, yeah. Yeah, whatever it is. There must have been just a few frames where the credit card was visible. The number was visible. All it takes. And that's all it takes. And I mean, somebody called it and thank you. I truly appreciate that. And it supports your standard. Look, I don't do this blindly. Like I know I do a lot of things that

46:58are quote unquote risky, but it's like part of what I do. So that credit card was with a bank that I have no other relations with. So it's not like my main banking, whatever. It's a very low limit card. I have alerts set up. You know what I mean? So it's like before ever putting it anywhere, like I knew like that, that thing could have been nuked and stolen. Like like it was set up with a firewall around a pandemic since to everybody that saw it and alerted me and talked about it. Thank you so much. I felt very, very silly. Out of curiosity, when you put your credit card number out on the internet, what do people try to buy? Yeah, do people are weird, man. Is it all computer components? Like new egg and stuff? Or was a let's see tons of developed related stuff. So there's like GitHub and this and that a bunch of charges for trying to get money out through PayPal. So actually getting trying to get cash.

47:59Few of them look like valentines gift select some sort of like you're like, I'm like, yeah, a bunch of sweet hearts in your audience trying to do the right thing to their loved ones. Well, I mean, I guess depends on what you mean by the right thing to PSA to the people out there if you're dealing credit card. Yeah, it doesn't make it okay. Just because you're buying flowers, doesn't mean that is like, no, you're doing valentines wrong. I think women will not appreciate that. Like it's still this credit card for you. It's not Aladdin. Yeah. So I lost zero dollars in zero cents and I'm pretty sure that the merchants, so because sometimes I used to run a lot of the e-commerce like with somebody charges stuff up and the person doesn't catch it soon enough. Um, the merchant can still lose money because if they shipped something to the person and then the dispute comes through, well, the merchant lost money, right, which is which really sucks. So in this case, I don't think anyone lost any money and I woke up and within two hours of the

49:04first fraudulent charges I was like canceling it. I was on the phone with the bank. But um, while I was on the phone, they're like, oh, did you try to call in earlier to report your credit card stolen? Somebody else had somebody else had and they mentioned in the class as well. So to that person, wow, you're an angel. Thank you so much. So if you find a credit card online, you can, yeah, you can call it and be like, hey, I saw this number online. You should cancel it. Apparently, yeah, the thing is at the time, I wasn't sure what to make of it. Plus, I wasn't really thinking about that because I was just trying to like, let's cancel this thing. I thought, okay, maybe that's like, what's it called? Social engineering? Somebody's like, you know what I mean? Maybe that was a way for them to try to get some information out. Like, oh, can you, you know, cancel this one, send the new one to this address. I didn't know what to think of it, but looking through the comments, I realized somebody did report it. So like I said, like, the good is amazing. People are amazing. Thank you so much. This kind of like, I still have a lot of trust in humans. There's always like the few bad apples that I think

50:06like ruined everything for everybody. I think the whole thing with open claw and mold book and stuff like that, that's really, there's probably just a few people that really did bad things and the rest of the people were nice and sincere and conscientious and all that. So I would think so. World's full of mostly good people. I do believe that very, very strongly. So still to this day. But yeah, so yeah, anyways, it was me being foolish, but it was like, look, it's like a stuntman jumping out of window. Is it risky? Yes, but it's kind of also you understand that this is kind of what the idea could happen. Yeah, this wasn't a shock to me. This wasn't unexpected. So I just do want people to understand that. But at this point as far, there's been no breaches. I guess as you say this with open claw, my use of it zero, zero security issues. I do not have it interact with anything where somebody can talk to it on the internet, right? So it doesn't answer email. You can read email. You can't answer email. It can read comments and it can't answer comments.

51:11You know what I mean? So that's the, I think the thing that kept me super safe. Just don't have anything where somebody can interact with it and give it instructions. Even then, I think there's a way to, you know, make sure it's secure, but I haven't touched that. I just don't have it interacting. And that's kept me safe and secure. And well, at least from the AI agents losing information, I still obviously lose love. Yeah, and something happened to your eyes. And now we know. Yes, yes. So this has been, this has been a crazy, crazy, crazy week for me. So I'm pretty comfortable with the idea that in the future, there's going to be a lot of self driving cars. And I can imagine like lots of cars similar to my Tesla, but with a future version of autopilot that is even more dependable. And I do know that in some cases, it will make a mistake. It might kill someone. It might hurt them. But I would think on average, it's going to be much safer than all the

52:11distractable people that are out there right now. It can, you know, see it, frame rates that are way beyond human vision. It has sort of cameras in all directions. There's abilities for it to sort of communicate and make predictions about what people are going to do before they do it and like avoid hitting them. So there's just so much positive potential. And even though there is some risk to me, it just seems worth it. And then when you take these things and put them in towards a war, I just get worried because like the point of war is not a goal that I am super comfortable with AI going for because I think that I think on the other side of this, like in another 20 years or 30 years, there could just be like so much abundance. I just, I could imagine robots just bringing the cost of almost everything to zero, having all sorts of like longevity. We could probably clean up the atmosphere and have all sorts of like beautiful parts of the world, like cleaned up and like able to live in like all that stuff is out there. Do you know what I mean? As long as we just like if we took it a little slower and like didn't build some of the things

53:12that are high risk, like I would like to just focus on those. And even though along the way, some things would happen, you know, hallucinations might happen that could be very dangerous. I just want, I don't like the idea of like a huge war machine sort of to actually be like intelligently moving in a direction that we have to then counter with like even more intelligent non-war machines constantly. But yeah. Yeah. There's a podcast called hardcore history Dan Carlin, I think his name is, um, he has one of them is about the first world war, like the best piece about like they've ever heard of, I mean, maybe on anything. So it's a podcast and I encourage everybody should listen to it. It's just kind of mind blowing. He, what he did there is truly, truly special. A lot of stuff there that I had no clue about. It's, you know, when you listen to it, it's almost like that war was like faded to happen. Because there was a lot of things that happened. It's like,

54:15like, like, what's that movie final destination? I think it was. It's like where no matter what you do, like, you're supposed to die and those people escaped death. But then all this like crazy stuff kept happening. You know what I'm talking about? I think. Yeah. Yeah. They just kept avoiding. It was going to find a way. It's going to find a way. So I think with that Duke Archduke Ferdinand or whatever his name is, friends Ferdinand, they were some like assassins after them after him. And they were supposed to find him on his, as he was like parading through the streets or whatever. But something went awry like the gun jammed or something happened. They missed it like some weird coincidence happened. And so these would be assassins there later that evening there at the pub and they're drinking and they're kind of like, like, I can't believe like we missed those were like our one opportunity. And they hear a car break down in front of the pub, right? It kind of like just breaks down and the driver comes down. He's fixing it. And they go out there. They look at it

55:18and guess who's in the car. It's the Duke. And that's how they got him. Right? And so and that's the thing that they say set off the war. You know what I mean? So it's like, yeah. That's strange. That's a strange coincidence. Well, you don't like I guess part of it is maybe me and you just sit here and like read this stuff so much. But it I just don't know if many people believe that it's like really coming. You know, I just think for some reason it's so easy to look at the past. And I don't blame anyone, especially if they're not reading the news like I started like thinking about how these models work or like what intelligence the sort of the nature of it is. But you know, if if all the militaries were like, Hey, there's there's a good chance that like we get through the singularity unscathed like a good 90% chance. But we're going to have to not militarize it. We need to like stop hurting each other. Like we just need to kind of work as an entire planet because on the other end is everything we want. Like what do you fight a war for? That's going to be

56:18any better than getting through the singularity safely. Like there's not like China could take over America and all they would get from us wouldn't hold a candle to what's potentially there if we work together. Yeah. Well, so I think the counterpoint, whether it's true or not. But the the counterpoint that exists, I think, is that a lot of people believe that stability is usually accomplished through an overwhelming force on one side. So whenever there's the enemy, that's kind of like what happened in World War One. Interesting. That's actually why I kind of brought it up because like the blood, blood bath was absolutely incredible because I think no one understood the true scale and the new machinery that was involved. So there were times when they were just like throwing these people cavalry against entrenched positions with machine guns. And it was like honor. You know what I mean? It was still like the old school kind of more medieval like,

57:20die for your country. And it was just waves and waves and waves and waves. And there was a couple guys years of machine guns. And then there was like thousands of people here on horses. And no amount of throwing bodies at the machine gun nest would ever take you know. So it's like one of those things where yeah, people for people that would know more about it, please because I might be getting some details wrong. Please, please like watch that. Listen to that podcast. It's absolutely incredible because it just spelled out how it was like the perfect storm of things going wrong because it was also like the population were populations of the countries were exploding because the the farming methods were better. So these countries were feeling bigger and bigger sort of amount of people that they could put out there was just unprecedented. He compared it to the wars of old. And it was like back then, the amount of people that died in the biggest worst possible is a joke compared to like on a chart, you know, World War One was just like way up there. Anyways, my point is that. Oh yeah, no good. You had a point. I didn't want to get you

58:25off on it. Yeah, no, just the point is that I think often they say that an overwhelming force on one side is what keeps the peace. Sometimes people also draw the connection because of what's happening like, you know, South of us in Mexico and South America with a lot of the the cartel activity and stuff like that. And again, I don't know too much about it, but my understanding is if there's one kind of overwhelming force, things tend to be a little bit more stable. When there's multiple forces that think they can take over, that's when like the really, really bad stuff. And you do see that in the animal kingdom too. I mean, a lot of like stags and chimpanzees like the fighting happens when two people look like they both could win. But when it's clear that one's a winner and when one bangs on their chest and the other one does and like one's clearly stronger and the other backs down, things are safer. But is that is that how you see it playing out? Do you think that either like China or some pentagon war machine comes up with some kind of AI model that's just so good at doing whatever the the control mechanism is like

59:32that it just ends up kind of saturating the world like everyone in the world just gets a notification that like, Hey, here's here's the rules of society. And if you break them, then these agents will come get you and they're decentralized and there's no way you're going to stop them. I mean, the scariest thing is both sides. And I guess I'm saying both sides as in US and China. Because even before AI, people were concerned because usually when there's like a world power has been around for a while in another world power, like is sort of ready to eclipse it or close to it, more usually that ends in conflict. You know what I mean? Just looking back at history. So a lot of people were very kind of concerned about that. So I mean, the sort of the scariest scenario is where both sides right here in the US, whoever's controlling the stuff, they're like, Oh, we're obviously have the overwhelming AI force. And they're in China. They're like, Oh, obviously we have the overwhelming AI force. And they're like, All right, let's pull the trigger, right? That's like the

1:00:33worst possible outcome I feel like. It's a yeah, just a quick chat GPD search was saying the US has about 5,500 data centers in China has about 500 data centers. So I guess we do have a lot more, but they clearly seem like they're better at robotics and kind of manufacturing at scale. So yeah, recently I was reading where they were saying that they're trying to build they're really like AI chip kind of piled in a sense that now they they they they know that they need to build the infrastructure. And here's the thing like if tomorrow the sort of global globalism goes away and everybody has to only survive on what they can produce in their own country. In that situation, China will be the de facto leader of AI chip reduction in the world. Now, there'll be like 20 years, you know, 10 to 20 years behind what's currently available. But they will be self-sufficient. Whereas for everybody else, including US, we rely on, you know, sourcing parts from all of the

1:01:36world to create like the greatest sort of world class technologies. But we rely on kind of was that globe not global is like like co-obitation or just logistics. Yeah, I guess it's yeah, just ability to source from all of the world. If if that goes down, China is self-sufficient and pretty much nobody else is I was talking about AI chip specifically. Yeah, it's wild. All right, well, let's see what this let's I think you wanted to show some things up before. We take too much of your time talking about war. What's well, I guess well, let's just let's just talk about it because I will be rolling it out. I feel like pretty soon here. So people will see it and chances are I'll probably set it to go live. So for people watching it, natural 20.com natural 2.0.com, that's my website. And in the back, I have a staging area where I'm building the new site. And when I say I'm building it, you know, open class building it, AI agents are building it. I don't write anything whether that's code or text or anything else. It had an idea. I told it go for it. And it's

1:02:39been building it out and it's been keeping track of it. And what it is is a news aggregator and still kind of being worked out. But so far, everything's looking really good. It sources the top news from a lot of different sources. It has a way of ranking those news in terms of how big the art specifically for AI. And also it's now writing up articles, a little little coverage here and there. I plan to integrate it with my newsletter so that it automatically creates the newsletter, sends it out. This is going to be a human loop on all of these. But my goal is to create a fully autonomous enterprise that kind of runs by itself. By itself meaning it's an asterisk. And I'll spell out what that means when I launch it. I'll kind of because there's some things that so far a human is needed for, especially a lot of things that have to do with payments. I'm not going to have this thing handle my payments for me. Or any, you know, user data or anything like that. But a lot of the work of the actual grant work will be handled one percent by AI's. So

1:03:46so far they've created a way to rank these things a semantic clustering so they can kind of group words, group stories together. And then also more recently just throughout this night overnight, what it was working on was figuring out how to cross link all the articles in a way that's going to make sense for Google and for SEO. One interesting thing is like I'm almost concerned with because what I want to do is I want to have this completely open and visible to the public like the actual how is being built and giving all the data and stuff like that. And one of the things that I want to get an experiment is SEO right with Google search engine optimization. And one of the ideas that I had is actually for each article listing where that article at the bottom, the AI just go through every day and they kind of show where that article is ranked in Google and Bing and whatever other search engines that are out there. By the

1:04:49way, my site is getting a lot of traffic from chat GPT interestingly. So that's one other thing that I'm going to be exploring like how does this they call it GE. Oh, it's going to ask for that. Yeah, like the SEO of what is it called? What is it? What does this stand for? Hey, they kept saying GE, which I'm like, gee, because I would have been like AI. Oh, yes. AI. Yeah, but that sounds, but no, gee, and I didn't, I'm like far was geo like geo, why geo, generative. I finally saw it today because the agent wrote it out. It's a generative engine optimization. So they're calling that generative engine optimization, which okay, I guess, I guess AI, EO sounds weird. It sounds like you're goreling. But yeah, so, but here's the thing. Let me, okay, let me, let me dive into this for like two minutes. It's really fast. So I used to do SEO back from like 2011, right? So I kind of followed the space and kind of follow a lot of the experiments and people posting their stuff. So basically for those that don't know, there's

1:05:51on page and off page SEO. And that's how, how much you rank in Google, right? Cause Google has ads, but the rest of the results are organic. And if you're ranking for the organic results, you get a lot of free traffic. And there's people that are making tons of money just ranking in search. And so how it works is there's on page and off page and on page is everything that you have on the page written out that tells the AI, the, the Google search crawl bots, as they're called, kind of what your site is about. And off page is just basically how sort of other sites recognize you. The more they link to you, the better it is, et cetera. And over time, over the last 10, 15 years, whatever it was, how Google does their sort of on page SEO, how they rank websites, more and more tools came out to try to like crack that algorithm, right? So over time, they, they realized that there's certain rules to it, obviously, right? But these sort of like reverse engineer that here's the thing, we're now in an era where you can have autonomous AI agents or

1:06:58sub agents, it's able to launch its own like sub agents to do certain specific tasks. That can run continuous experiments on these websites on each page, shifting how the worst structure is. And then waiting 24 hours or whatever, seeing how that affects the movement of the actual listings on, you know, in Bing, in Google, et cetera. And just updating it. And the thing is it's not like rocket science, too, because like if the number one result, if you're just copying it, or you're looking at what's working there, right? So it's not like it has to understand everything. It's like, okay, your number five, it's number one, what is the number one thing that you're doing that, that you're not doing? What are they doing that you're not doing? And so it shifts it. And that can be running on every page. And that's, you know what I mean? And it could be out there, like reading all the other ones that are above you coming up with theories on why they're doing

1:07:58better, reading about the algorithm, then updating your text over the evening. And the next day, that's crazy. So what do you think Google would do if they saw somebody just putting all that stuff out there as an open case study? You know what I mean? Like, they might not like it, because that seems like a pretty big threat to the like, but what's the worst that's going to happen? They're going to like, do you rank you or something and like you have your own audience to share with? I don't know. Like, it's interesting. I don't think you would get anything to serious. Here's a funny story that I think some of the people are going to get a kick out of back in the days. This was a while back, 2016, like 10, 10 years ago, there was a person I used to know his name. He, there was a SEO competition. So people basically had to all create brand new sites, and whoever was able to rank and get the most amount of traffic from Google would win. So it's like kind of this SEO nerds kind of having a little fun contest, whatever. One person who's well known in the history for kind of like knowing how to do this stuff

1:09:00creates a website. It ranks. It's pulling in tons of traffic. Here's the problem. He used the Lorum Ipsum text on most of it. So the headlines, H1, H2 tags, like the headlines were keyword rich text. So if it was like, you know, best workout gear, it would say best workout gear. In a paragraph, but the paragraph was gibberish. It was Lorum Ipsum, right? Lorum Ipsum is that nonsense thing that looks like it's English. So he ranked a gibberish website just with, you know, keyword stuff, keyword stuffing, I guess you can say. Yeah. I remember back in the day. Now that site a week after it was like kind of published that case that it was published. That site got nuked and any other sites that he had related to that site just got nuked. Like Google just dropped the bomb on them, right? Because they don't want people knowing that because that's like, because most likely they have some, yeah, they could flag you and you know,

1:10:01you also have a YouTube account to worry about and all sorts of other stuff. So there's stuff that I would love to test out that I'm not touching. I'll mention it. Like here on this podcast for people that are interested in maybe testing it out. By the way, if this is some that you're interested in, you know, maybe like leave a comment. Let me know. Let us know that you're doing this because we'd love to kind of maybe pay attention to it. And if you yeah. And if you wanted covered anonymously, like if you want to talk about the case study and you just don't want us mentioning the actual website and stuff like that, we'd be happy to do that as well. Because this seems like it's going to be massively disruptive. But yeah. So I mean, it seems to me like the algorithm when ranking pages on the internet would notice new pages and rank them down. I mean, if you were sent up for like a new YouTube account or a new Instagram account, it's like very low trust. Like they don't let you just post very often and they're not sure who you are.

1:11:03And then over time, other people kind of interact with your account and the trust score goes up and you get more kind of freedom. So it seems to me like you might hurt yourself changing the content every day because it's going to be like, oh, this is new. This is new. This is new. And then it's really at least the way I used to understand the algorithm for ranking. This was years ago, but they really didn't trust what was on your page very much. They trusted who linked to you. But if you have the domain rating, then they trust each new thing that you published. And I think it was 2012, they released a caffeine update, which actually gave positive signals for freshness of the page. So a lot of people just game did by making daily updates, especially on certain subjects. So, you know, new new subjects needed to be updated more often versus, yeah, if you're doing something historical, the history related, like maybe you don't need to update your world word to wiki pages that need to be updated every day, because there's no breaking

1:12:04news about that usually. Whereas, you know, some of this news related gets updated constantly. So yeah, it's complicated, man. There's a lot out there, but yeah, it'll be interesting to see if anyone does that as an experiment. I'm curious how much of a disruption stuff like this causes. I feel like it's coming, man, I can't I predict more disruption all every time. That's a good bet to win the one that I can make. Yeah. So for people that are interested, so it's not live yet, but I plan to launch it live this week. Probably not going to be doing a huge fanfare on day one, because again, once like humans make contact with it, I'm sure there's going to be tons of stuff that's breaking. Oh, there's issues. So I'm probably going to do a stealth launch and then just kind of see what happens and then kind of announce it and go from there. So if you're interested in this idea of autonomous enterprises, by the way, I'm not promising anything because I know there's like, I'm sure there's people selling courses and God knows what like, oh, make millions in your sleep while AI agents do the work.

1:13:08If they don't have that out there, trust me, it's coming. You know, I don't want to be that guy, but at the same time, it is, this seems like a big deal and kind of keeping and learning how to do it is going to be. It's crazy thought just because I was like, if you did it, you'd have moldbot do it, which would be the most interesting article ever because if so, if I got advertised a course on how to make money and like, you can trust me, I'm an AI agent myself, I'm not even a human. If you want to know how to use AI agents to like, make more money by my course, like, that's like how they're going to make money. And we're going to trust them more than a human to write the content because we're trying to use them to make more money. Okay. Is there a light bubble in my head? No, this is, this is, well, I guess it's your idea. But no, you're absolutely right because really, what is my sort of, yeah, you're just going to slow it down there. You're the human being like, go to WordPress and it's the AI. It should write the article. Yeah, I, let me know

1:14:12the comments. People, what do you think about that? Like, let's say, let's say I do manage to get this autonomous enterprise. And again, I'm not promising. I'm not saying it's real. I'm not saying it might, I don't want to say fall fall on its face because I don't want to use that terminology right now. But it might fail. It might fail and it might not work very well. And that's very possible. So again, what are the chances of it making tons of like not even tons of money? What is the chances of it being profitable? Let's say this year and pulling in, let's say a few thousand. I just think if you, if you, if you have AI write a course about how to use AI to make money, you should give the AI the actual profits so that it can respend them. And so I guess that's my question for people. Would you prefer that it's a human being that's writing it? Would you prefer it to an AI that's writing it and just cataloging everything that it's doing with maybe, you know, like a human loop making sure that it's correct and stuff like that. Would you be okay learning

1:15:17from an AI? Assuming it's like real time up to date, it's actually building the stuff and then it's giving you the actual data. How do you feel about that? Let me let us know in the comments. I'm kind of curious. What about you? Would you be okay with that? Well, I mean, I'm certainly leaning on Chad Gbt for a lot of replies. And I go to ontropic in Google all the time. I mean, even when you were talking about search ranking, I was like, it don't, once less time, I use Google to just click on the top link. I mean, I use it to find websites, but I don't really search for knowledge using Google anymore, you know? Right. It's more of a, it's more like a database than it. I mean, I remember five years ago, I used it to actually find knowledge, but yeah, I'd never, just hit Wikipedia pages naturally anymore and whatever else I used to just like people, random people explaining how to do things. I'm always on a generative model. Yeah, this is kind of blowing my mind because I think about it. Like, let's say this thing builds out

1:16:17in AI news aggregator, all the stuff that it did, all of the learning, all the data, all of the things that it tested and tried and whether failed or not, all those are valid data points. Let's say somebody else wants to create something very similar in a different industry. And there's like a million industries, right? There's niches rather, right? Having this AI chatbot could translate that and give full guides, not just to the person, but in fact, to the actual agent of that other person, for him to hammer her to start building, you know, kind of like a, not a replica, but kind of take that idea, take those findings, that concept and kind of rebuild in a different industry. Yeah, you know, maybe you're like, maybe you should tell your agent right now, like, hey, record, record everything that you're doing because one day there's going to be a documentary made about you and, you know, document like all the big steps, the milestones, the setbacks, like, build a narrative around what you're doing. And then just keep that in some folder. And then

1:17:20later you can just like point to that folder and be like, use some sort of future version of Sora, like make me a documentary. So I do have that. You're the agent. And it's on the website. So that's going to be live and public. Okay. But maybe just, but those are summary. That's like a summary. You're talking about more in debt. I don't know. Yeah. Like it could be like a tutorial, or it could just, I don't know. It's just like whenever a big company like goes from nothing to something, like just if somebody was following it and making a documentary, I just could imagine. Let's make a timestamp right now. So this is February 17th, 2026. Let's fast forward to just before Christmas, December 17th. Well, this will be how the movie starts. This is how the movie starts. Like, what's your gut feeling about either this sort of business or just businesses run by AI agents in general that some other people start, like, you know, it's December. Do we go, wow, we were so foolish that we believe that AI agents could do any of this stuff. Or do you think we're like, wow, that was a crazy time when this whole thing took off. Like, yeah, I think it's

1:18:24crazy time. Yeah, I mean, you just, you just kind of know like, like you, like, remember what blockchain just kind of was out there and you're like, oh, I see why this is different than the traditional financial system. Or like certain things with biotech, you can kind of feel they're like on the verge right now. Like to me, this feels like it's on the verge right now. It just, yeah, I mean, it seems kind of obviously clear that the like newspaper industry was going to get disrupted because it's like so convenient with a Kindle or a digital format. And to me, I have a business and like, I think about how many times I like just go into folders and pull up numbers from some document and then send them to some lawyer and just none of that stuff really needs to truly be a human. It's not creative. It's not there. And that's the backbone of these businesses. The ideas at the top are almost some of the easiest parts and all the execution is the hard part. And this can absolutely execute. So why wouldn't it build a big business in the near future? Yeah. And all the stuff that I'm running into the issues, it seems so silly because it's

1:19:24like, it's brilliant at building everything out. And it's like, oh, this web page is up. Check it out. I'm like, it's not up. It just says, you know, 404 or whatever. And you know, I tell it's like, no, it's live. I'm like, okay, send me a screenshot. It's like, here you go. And it's like, yeah, okay, look at the screenshot. It says 404 on it. He's like, oh, you're absolutely right. And then it, it, it troubleshoots it. And I have no, like, what it's saying, like, oh, there's a stale, casheed version of that and I have to push it to resell and redeploy to get, I'm like, fine, I don't need to know. Just do it. And then a minute later does it. So it's like really like, you kind of like dirt out at, you don't know what, you know, you can't take a screenshot of a website. That's, that's where like, by human intellect has to step in, but you, you know, all the stuff. Anyways, but I think let's get fixed. Yeah, it is. So we're like, one time my friend Dan was putting a, he like had a picture and let's say it's a dinosaur or something and he's like, put a hat on the dinosaur and then it was like, no problem. And then it came back with like a picture of like a cup. And I was like, I asked you to put like a hat on a dinosaur and you gave

1:20:28me a picture of a cup. And he's like, you're absolutely right. I don't know where that came from. And then like the hat with the dinosaur comes next. I'm like, where'd you get the cup from? Like, there's like, you weren't even in the ballpark, you know? Yeah. Yeah, it's so interesting. It's like, and I, I commonly now run into that where it's like, it's just a stream of geniuses, genius, you know, flow followed by just like this like, what dirt thing? Yeah, dirt thing out. What would you say? No, that's not what I meant. No, redo the whole thing, especially when you just, he just wrote back WTF. And it was like, you're right. Like I said, you a picture of a cup, a US for a hat to be added to the original dinosaur. There you go. Like, so, so off, you know, like if somebody, if a human was that far off, you'd be like, what? Yeah. Yeah. Exactly. But I feel like that's going to, well, so first of all, if it, if that is interesting, it never gets fixed. First of all, this whole thing would be hilarious. Number one, number two, you wouldn't really have to worry about AI safety as much. I mean, can you imagine this like super intelligence that just, you know, gets like, you're right.

1:21:31That'll be, that's what, what you need is it called the Derp benchmark. And it's just like, does, does this thing not derp out after one trillion tokens? Yeah, exactly. We have super intelligence. Because what if the, for whatever reason that's just built in? Like it's a genius, but it can't ever carry out a full-fledged long horizon task without doing something like incredibly, like, like, like a child could fix, you know what I mean? Maybe. I mean, I will say, I don't know how long that horizon will take. That's one of the horizons that doesn't look like it's getting solved like this month or next month. Like it might, I could see a fundamental change I didn't see coming or I could see that taking several more years, you know? It does seem like that's a tricky part that I haven't seen solved yet. Yes, there hasn't been, oh, I mean, so Google has been talking about their long-term stuff. But we haven't seen anything that's like, we've seen the first paper published. We haven't seen any sort of results. Usually the way I say it is like, you see the paper and then six to 12 months later, you see a commercial lab actually, you see it in action. So that's only happened, I think, late last year. So maybe this year,

1:22:36we'll see somebody actually figure out how to deploy that live. So did you dive into deep think the Gemini 3 update? No, there's been so much stuff happening that I haven't. But I know it nailed some crazy benchmarks on bench proof math or physics or something like that. Do you know anything about that or anything in general about that? Well, all I know is that when I look at a chart of benchmarks, some of the ones that nobody's ever done good at, it did extremely well add, but it did seem like it almost came at the cost. Like the other ones didn't budge if not went down. So I know they tried. I know that model has kind of a new way that it's kind of going about solving scientific problems. But I get talking more than that right now. I didn't look into it enough. So yeah, neither have I unfortunately. It's just like, it's like you can't keep up with everything at this point. It's going way too fast. So yeah. But all right, we should probably end up just poor people have been listening to us

1:23:36ramble for a long time. So thanks for everybody hanging on and we will catch you guys in the next video. If you made this far, thank you so much for watching. You're a trooper, amazing person, and we'll see you in the next one.