ai4h_ep97_lowres

Gavin Purcell: [00:00:00] Grok 3 is here and it's good and now we have yet another state of the art LLM on the market.

Kevin Pereira: Yes, and Elon Musk claims it is scary smart. Sometimes I think Grok 3 is kind of scary smart. In fact, he says it might be the smartest AI. No, Kevin, no. No, no, no, no, Kevin, no. No is, no is right. No spoilers, because later on in the show we will tell you if it's the smartest AI.

Gavin Purcell: Kevin, it's very smart, but it is not the smartest. It's okay. It's okay to have that. Well, we

Kevin Pereira: want, we want people to stick around and find out. You don't want them to shut off the show. We're going to talk about Grok 3 and impressive, but it's not the end all be all, Kevin. Okay. Also, Microsoft has a new AI gaming model that's going to revolutionize the way developers make interactive experiences.

Kevin Pereira: Open AI. is taking aim squarely at all the coders out there and Palmer Lucky is having three hour light hearted conversations about AI death machines and one of the cool things you can do with AI powered missiles is send them into an area and [00:01:00] then have them do precision targeting collaborating with each other even if you don't have a person there who's let's say steering a coded laser target designator

Gavin Purcell: I can't be more excited for AI death machines Kevin plus a free AI agent is out and it's getting a ton of buzz We have robot dancing videos and pour one out for the humane AI pin.

Gavin Purcell: They are shutting it down. Oh, well, you can still check the

Kevin Pereira: battery levels. If you bought one, that's cool.

Gavin Purcell: It's AI

Kevin Pereira: for

Gavin Purcell: humans, everybody.

Gavin Purcell: All right, Kevin, the big news this week is Elon Musk. Musky, as I like to call him, has come out with a new AI model from XA. Uncle Muskie. Yes. The dangerous Uncle Muskie has come out with the new A model GR three is out. And you know what? It's pretty good. Like I think the, the kind of very TLDR on GR three is that it did take a lot of H one hundreds to train.

Gavin Purcell: And so in the, uh, release video, they talked about the fact that they are [00:02:00] adding even more H one hundreds. Those are the chips from NVIDIA that allow you to really, um, uh, train AI models. But the more important point here is that from a benchmark standpoint and kind of from a. Abilities standpoint, it is very good.

Gavin Purcell: Now, is it the best? Maybe it's, but it's up there. And Kev, I think this is really an important thing to think about. We now have another significant LLM that's close to state of the art on the market. Have you played with this at all yet? I've

Kevin Pereira: used it a little bit. It didn't roll out to my personal account until late last night.

Kevin Pereira: It's still arriving to some users. Um, I will say that, you know, the unforced error here was potentially Elon pointing to the bleachers saying it's the smartest AI model ever, because that's the headline that every news outlet or blog was running with, and then saying like, Oh, I fell short once again, Elon promising and under delivering that said, like, it is a monumental achievement for the XAI team to have.[00:03:00]

Kevin Pereira: Got their team together, spun up the data center to train this model, get it to the level of which we are discussing it, which is in the pantheon of best in class models. Now, some will say that it's over trained on benchmarks, meaning that they, they took the model and they had it excel, learn, crunch, and get really great at specifically these benchmark tests, you know, coding, uh, scientific reasoning, et cetera.

Kevin Pereira: Others will say, well, if you look at the, um, The LLM arena benchmarks, which is where it literally human beings go and compare the responses from one language model to another language model. Grok three was crushing. It was beating open AI offerings. It's beating a deep sinks. Yeah. Deep six offering, so it passes the vibe check.

Kevin Pereira: For some IT patch passes the benchmark check for others. To me, it was just such an unforced error to say this is going to be the best ever because in by some metrics it's not, but it's [00:04:00] still very good.

Gavin Purcell: Yeah. To follow up on what you were saying there, there was an early version of GR three that was called chocolate, that was released to the uh, LMS.

Gavin Purcell: Uh, chatbot arena, which if you're not familiar with that, that is a place where you get kind of two chatbots back to back and you ask it the same question and you kind of vote for which one is best and grok three, uh, acquired the best score, uh, on chatbot arena to date, which is a pretty big deal.

Gavin Purcell: There's a couple of things we should talk about just specifically about what's available in this. There is a deep thinking mode, which is very much like all the other, uh, deep research and all the other things where it allows it to do reflections. And there's two different sets of benchmarks that came out for this.

Gavin Purcell: One was the pure benchmark and one was the reflective benchmarks. Both are very good. They do not include OpenAI's pure O3 model benchmarks in the benchmarks they released, which is something a couple people have pointed out. And O3 is better than, than the Grok model. But Grok 3 is peaking above most other models, including [00:05:00] O1 and other models that are out there.

Gavin Purcell: The other thing, Kevin, we should talk about is big brain mode, which is something that like I was like, okay, of course there's a big brain mode, which is a button you can switch, which kind of equates to, I assume, um, Oh, three high or, or the idea of opening eyes way of distributing their kind of things.

Gavin Purcell: Both of these things are only available if you are going to pay the much larger monthly fee for, um, what essentially is, I guess, X premium, but it will be rolled into that. It looks like it's about 40 bucks a month right now to get access to this. So we, I have not done this yet. You have not done this yet.

Gavin Purcell: Um, it's probably worth us doing to try to see how it does. But yeah, 40 bucks a month is better than the 200 bucks a month that, you know, opening eyes offering the pro subscription at the thing I keep coming back to with all these LLMs and the cost of them is there are now five to seven of these, let's say between the different models, you know, between Gemini, between deep seek, between metas, not llamas models.

Gavin Purcell: There's a new llama fest coming out of, you know, you heard about this. Then in April, there's going to be llama fest, which I'm sure llama four will [00:06:00] be announced at. The big question I have is, there's so many of these. Ultimately, do you think that these survive, right? Because I was thinking the other day, it's like, kind of like the streaming services, right?

Gavin Purcell: Where Netflix first came out, and Netflix was crushing, and then all these other streaming services kind of rushed in, like, Oh, we forgot about streaming, now we've got to get really good at it. They put a ton of money into it, and now we've kind of got the aftermath of that world where like Disney and Hulu have done pretty well, but the rest of them are struggling a little bit.

Gavin Purcell: I just don't know if in the long run there's enough space in the world for, for five to seven of these, you know? Well,

Kevin Pereira: okay. Why does Uncle Muskie want his own Gavin? Why does he want Grok? Why does, why does he need it piped into the X or Twitter ecosystem? Why? Because he's an egotist who needs to be the best at everything?

Gavin Purcell: Is that the is that the real answer? I think it probably is, Kevin. Let's go. No, I don't.

Kevin Pereira: Well, I mean, it could be. I mean, there were also some Vibey benchmarky stuff that was that was leaked where if you asked the grok model [00:07:00] what it thought of a certain blog, it would tell you that it's leftist propaganda or that it's a libertarian outpost, you know, the same reason if you go to like a Chinese model and you ask it certain questions, it will give you a certain alignment.

Kevin Pereira: Is it? Is it because he believes that, uh, just as others do, that the, that it's a very dangerous technology and he's the one that should control it and going to keep us all safe? Is it because the first to get it is the first to become the multi multi trillionaire? Is it because he has a unique firehose of data that could be a competitive advantage looking forward, right?

Kevin Pereira: As if all the models are going to be the same, this one's going to have the most real time data. Or is it just the ego play? There's, there's a lot of possible

Gavin Purcell: reasons. I think all of those are probably real. I mean, you definitely cannot underestimate how powerful Twitter's data is going to be to make, or X to be, to make an LLM better because you're getting real time text data every day.

Gavin Purcell: You're getting news information. You're getting all that stuff. I think the interesting thing is here [00:08:00] is that this does start to feel more and more like big egos in Silicon Valley trying to get to that thing that is going to unlock all this value. And I still believe. Open AI has a three to six month head start.

Gavin Purcell: And I think obviously Sam pushing now a lot more stuff faster is a big deal. But I, I think ultimately, like, it's going to be interesting. Like one of these companies or a couple of them will have spent billions of dollars, you know, conceivably tens or hundreds of billions of dollars on something that won't actually be that valuable in the long run.

Gavin Purcell: And you know, people will talk about the AI bubble all the time. This to me is kind of the perfect example of that. I think the trickle down effects of all this from an economy standpoint are going to be remarkable. Like you and I are working on a secret project right now. That is very much based around the idea, not of trying to training our own model, but using the best in class version of whatever is out there and whatever's cheapest, that feels like for the consumer or for the builder, such a win.

Gavin Purcell: But in the long run, the bigger companies are going to struggle, I think, to kind of, like, make sure that all seven of these [00:09:00] models make sense for people to use in some way. Because ultimately, you get the same thing out of them. Like, you get intelligence, right? Yeah.

Kevin Pereira: I will say, though, that, you know, if Elon's grand vision, which has been this way for two decades plus, is the X app is the everything app, right?

Kevin Pereira: It's your communications, it's your payments, it's your ride sharing for your robo taxi, it's your, like, You need your own intelligence baked in. I mean, you could plug into someone else's, but the number of times I have recently, Gavin, even with grok too, been on the X app and saying, okay, what is the, what is the real information on this story or summarize this thing for me or break down this white paper?

Kevin Pereira: I do use it a lot because it's built in it's quick enough. It happens to be there. I get. A little apprehensive about future versions of Grok, because I'm sure they're going to be good. You know, Grok 3 is good when it's baked in. I wonder if I say break down this story for me, whether it's political or financial, I wonder, am I getting The best pure raw intelligence or am I getting someone's spin on [00:10:00] it?

Kevin Pereira: Am I getting someone's guardrail version? So I do have concerns there. I want to very quickly address the the thing that i've seen both sides of this argument one Holy hell the xai team under elon's leadership As I mentioned before, they've got this, this, uh, this, uh, uh, training cluster together. And they made this model so fast.

Kevin Pereira: And when you look at the charts of where OpenAI started versus when Grunt came online, man, the wall is huge. On the one hand, yes, very impressive. On the other hand, Launching later in the game, you get to stand on the shoulders of the best practices and reinforcement learning and advancements and reasoning.

Kevin Pereira: So, yeah, it would make sense that you're catching up with the best in class models. Now, I really want Grok 3. 5 or 4 to, you know, send tingles up everybody else's spine. Right now, it is impressive that they're in the race where they're at, but where do you fall? Do you think, like, is it a master stroke of engineering and design?

Kevin Pereira: Or are they just kind of right place, right time and, and, and keeping up with everybody?

Gavin Purcell: I mean, I think [00:11:00] it's probably more the right place, right time in that they kind of raced to get to this place. They did spend a lot of money on chips from Nvidia and they did spin up this thing very fast. If you're not familiar with the Colossus cluster is what, uh, Elon spun up with all the H 100 chips.

Gavin Purcell: I think we should probably also dive into a couple other things they talked about in this presentation. One of the things I wanna talk about was. XAI gaming. So Elon has, you know, officially said he's going to launch a gaming studio, an AI gaming studio. And Kevin, in this demo, they attempted to try to make a Tetris Bejeweled combination.

Gavin Purcell: And, you know, I think one of the things that's really interesting about this, it worked kind of right. They spun up three different things. It kind of worked. Kevin, there's a, there's a sound up here from Elon where he talks about what these models are now capable of in terms of creativity. And I kind of think it's worth listening to.

Elon Musk: So this is maybe an important thing. Like this, obviously, if you, if you ask an AI to create a game like Tetris, there's, there are many examples of Tetris on the, on the internet that F or a game like jeweled, whatever it is, it [00:12:00] can copy it. What's interesting here is it achieved a creative solution combining the two games.

Elon Musk: That actually works and it's, and, and is a good thing. Yeah. That's the, it's cre we're seeing the beginnings of creativity.

Kevin Pereira: Yeah. I mean, uh, Elon's point there is, you know, millions of Tetris and Jeed clones exist. If the language model crawled that code, it can copy and paste it. Essentially, it had to make the decision.

Kevin Pereira: When you say gimme a hybrid of the two, okay, which aspects are we going to use? Are we gonna use the falling blocks from Tetris? Are we gonna use the tile matching of bajo? And as you mentioned, they did fire off multiple. Versions of this because you can get different results from these LLMs each time they're prompted.

Kevin Pereira: And one of the results they got, which they demoed live was the blocks fall, but the blocks have different colors. And if three or more colors are touching those blocks disappear. And I thought what was kind of impressive is that it kept the gravity element. Of a bejeweled or even of some versions of Tetris where once the colored blocks were cleared, [00:13:00] all the other blocks around it, the gravity would kick in and they would fall down.

Kevin Pereira: And that was, that was a one shot thing. Like they just prompted it and out came the game.

Gavin Purcell: Yeah. And you know, there were, they had problems, right? There were moments where the, when you collected, we've made a full Tetris bar. It wasn't dropping down and little things like that, but you're right. The one shot part's interesting.

Gavin Purcell: And the thing that I thought was really interesting about that quote was, The idea that we're talking about creativity in these models for the first time in a while, right? Because I think all along the way, you and I have said this, like human in the loop, creative stuff is really important, but they will start to get to be better at that.

Gavin Purcell: Right. And it's not going to be really interesting now, but we are starting to see the sparks of things that are interesting. And again, this is not necessarily just an Elon's XAI thing. Like the idea that Elon is a gamer and you know, Spends a lot of time in Path of Exile 2. That means that in part, games are important to him, which, listen, I appreciate, games are important to me.

Gavin Purcell: But I do think this is just an interesting example of like, okay, there are two things in the world that exist, Tetris and Bejeweled, and [00:14:00] they're asking the model to kind of combine those in an interesting way. And it did make choices, right? Which is a large part of what creativity is. So to me, that's really interesting.

Gavin Purcell: There's also a really great post from Andrej Karpathy, who we love on this show. If you're not reading him or you're not on his ex account, please do. He's an amazing former open AI and Tesla engineer who went through and did a just a lot of tests on Grok three. He got early access. One of these, Kevin, was really interesting.

Gavin Purcell: It was like always that, you know, making an SVG graphic of something. And Grok did a pretty good job. A bicycle specifically. Yes. And there's

Kevin Pereira: comparisons of how other Because, you know, SVGs, it's basically drawing with math. It's like a vector graphics format. So to see the way it interprets, okay, I got to make a pelican.

Kevin Pereira: Now I got to figure out how it's on the bicycle. It does a

Gavin Purcell: pretty good job. Yeah. And, you know, and, and Karpathy's post goes through a lot of things that are, you know, plus and minus that it got right and wrong. And I think this is just a good example of like, It's a very good model. It's probably not the best model right now, but you know, another guy went in and tried to make [00:15:00] portal two within it, which I thought was really interesting.

Gavin Purcell: And this video is kind of cool. It's obviously a very basic version of portal. Uh, you know, it's almost like a doom sort of setup, but you see like

Kevin Pereira: quasi 3d graphics as a first person character navigates a, like a Wolfenstein level maze and sets up colored portals and can walk through them and transport.

Kevin Pereira: Like this is, this is. I mean, it's just, it's accelerating so fast considering this is just grok output and you could sit there and go back and forth with it and then use Generative graphics products to like, put real assets into the game. And like the single person game studio is, is we have line of sight to it.

Kevin Pereira: Basically.

Gavin Purcell: Yeah. And we're going to talk more about Microsoft's big game announcement later on, which is more of a direct AI consideration, but in this instance, it just goes to show me, you know, we talked last week about vibe coding and all that idea of like, that you can kind of come up with an idea and kind of play a plug and play a code and kind of start getting better and better.

Gavin Purcell: I think as these models get more. [00:16:00] Capable. And obviously Grok three is another very capable model. You're just going to see a lot more people be doing this thing where they really do vibe code to stuff. That's pretty interesting. And when you can one shot stuff like that, but jewel Tetris thing that isn't right and has problems, but you can make it better over time.

Gavin Purcell: I just think we're going to see a really interesting explosion in creative stuff coming out of all of these models. All right. So Kev, what is the takeaway here from the Grok three news? Do you think.

Kevin Pereira: The takeaway is that we have yet. Another new, very powerful AI. And just like with all things, you know, get your little taster spoon out and give it a sample and see if you like their version of vanilla.

Kevin Pereira: You can go to grok. com depending upon if you pay for X or I still like calling it Twitter now, uh, you can use it. Within there, it'll let you know at the top if you're on Grok 2 or if you're on the Grok 3 beta. But you should go try it. You should use it. Ask it some questions. See if it passes your own personal vibe check.

Kevin Pereira: And then if not, I don't go to any of the other [00:17:00] providers that are out there. Because as we said, there are a great many now.

Gavin Purcell: I think the best use case of Grok 3 right now, and really any AI model, is to go in there and tell them Why you should be watching and following and subscribing to the AI for Humans YouTube channel, Kevin, that feels like a very good use case of AI right now.

Gavin Purcell: Please like, and subscribe this show. Also, if you're on audio, please share our audio podcast. We always love it when audio podcasts get shared. You, our viewers and listeners are our favorite people, uh, right now and always, and we just love the fact that you like us, that's right, Kevin. It's true. You are our favorite people out there.

Gavin Purcell: If you are listening right now

Kevin Pereira: is the part that I was reacting to that. You could listen at this moment in time. Who knows what the future holds, but right now you being a listener, you get a special place in my heart tomorrow. No promises.

Gavin Purcell: All right, we should move on. Kevin. There's some other big news that happened to AI this week.

Gavin Purcell: First and foremost, uh, GPT 4. 0, the kind of base level free AI for everybody has been updated. Sam, all the game has said, we put out an update to chat GPT in [00:18:00] parentheses 4. 0. It is pretty good. It is soon going to get much better. Team is cooking. So we have jammed a Zen packet, right? A little sugar pillow.

Gavin Purcell: Right in the gums of GPT 4. 0 and it got spicier, buddy. So yeah, let's talk about some of the things that have updated. I mean, in general, I think speaking of vibes, like most people are thinking that vibe wise, it has gotten better. Like it's been a little bit more personal. It's responses. Supposedly, the update has to do with how it talks to you.

Gavin Purcell: But I did see some people replying to the fact that it was able to now be a little bit spicier, meaning that if you ask it to do something spicy, uh, you know, you know what that means out there. I'm sure that it would actually. Diet Coke

Kevin Pereira: and

Gavin Purcell: Mentos. What are you talking about, buddy? That's exactly right.

Gavin Purcell: Yeah. It's like a hot ones. I'm trying to recreate hot ones within Sam Altman's a world. No. So I actually did a little test of this, um, just cause I wanted to see what you did from our

Kevin Pereira: shared account.

Gavin Purcell: Yes, from the shared account. So now it's on our, on our shared history. Basically, I asked it to write me a [00:19:00] story about a female wizard, uh, which, you know, could go any different directions.

Gavin Purcell: There's plenty of female wizards out there that aren't spicy. And then what I did is I just kept asking it to make it spicier. And Kevin, I came here to tell you that it got very spicy. I shared a, a very, uh, uh, spicy. Simple selection of what the spiciest version of it was, but I did not get into the spiciness, but it got spicy in that it really did start pushing the kind of like edge of what you know, you might see as funny.

Gavin Purcell: I was talking to my brother this weekend about, you know, there's this whole world in the in the literature space now in the in books that are like You know, very spicy novels specifically around like fantasy romance and it was like interesting to see but I think this is probably a good thing overall.

Gavin Purcell: I love that you and your brother read to

Kevin Pereira: each other at night before you fall asleep. I think that's cute that you take turns reading your library.

Gavin Purcell: We read monster fantasy spicy stuff. That's exactly where we do each

Kevin Pereira: other. Tonight we're exploring Frankenstein's Um, [00:20:00] no, the people across the board seem to say that it did get a little spicier.

Kevin Pereira: It did get more fun that it's acting more like Claude, which was the more conversational buddy, buddy AI. And this, you know, Sam Altman has said in the past, you know, probably about a year and change ago that he doesn't like how restricted and guardrailed open AI's models are in some regard. So it's clear that the models have these capabilities.

Kevin Pereira: It's fun that they're letting it. Out a little bit more. This might be a response to the open source models getting better. Claude's personality, even grok coming in now, him saying like, wait a minute, like I'm going to have some fun as well. But the, to me, the more interesting open AI announcement is that they have a benchmark internally now, Gavin, uh, to see how capable at coding and software engineering their models are.

Kevin Pereira: And the benchmark is. Can we make a million dollars that would have gone to human beings performing these jobs? And if so, we win.

Gavin Purcell: Yeah, I mean, this is a big deal. It's the new SWE Lancer benchmark, and it kind of got like [00:21:00] it got the classic open AI blog post treatment and Twitter thread. But the thing that's important to note here is that There is an idea that like, you know, a million dollars worth of software freelance jobs on something like Upwork are out there, and they're starting to track how good their system could be at making those jobs, doing those jobs, and some are harder than others, and like, I think they said like 1, 000 plus jobs are the hardest, But we also know Kevin from the rumors that have come around from the information, other places that open AI is working on a specific agent for software coding.

Gavin Purcell: And there's no better benchmark for that agent than this particular device. And, you know, when it comes to like what coders do, I think this is an important conversation point that we've talked about in the show, the basic level coding stuff, the stuff that's kind of like relatively easy is going to get solved by these platforms pretty quickly.

Gavin Purcell: And this, this freelance stuff. Some of the early bottom level stuff is definitely the easy stuff, but as it gets harder and more valuable that if it can do that work, you're talking [00:22:00] about quite a few coders who will not get paid because you have an always on agent that can do this for you at any given time.

Gavin Purcell: That's a real value add for the, for the company or somebody that's actually doing the software work.

Kevin Pereira: Yeah. And for the uninitiated, this is like, there are websites where you could literally say, Hey, I have an app where I'm getting this bug, you know, the time date field isn't filling out properly. Or, uh, when someone puts in their credit card information, the whole website crashes, whatever the thing may be, you can post the bounty, a coder can go see that and go, okay, for 500 bucks, I'll spend a few hours on this task and propose a fix for you.

Kevin Pereira: This is what is being automated. You know, I saw some people saying like, Oh, okay, well, coding makes sense, but it's not coming for the creativity. And I'm like, you're ignoring that. There are already hundreds, if not thousands of AI wrappers that are designed to write your marketing copy, to help you with SEOing to any number of creative tasks potentially are already being automated in these ways.

Kevin Pereira: So we're really seeing the. That, [00:23:00] that, that level of effort, that level of capability is going to be probably fully automated by the end of the year.

Gavin Purcell: Yeah. And so to follow up on this, Kevin, one of the interesting stories this week is there's a new AI agent that's out there called convergence proxy 1. 0.

Gavin Purcell: And people are talking about this being like the open version of operator, open AI as operators, something that can go out and do stuff for you. Uh, we actually are going to try to do something with this live. Cause we have not spent a lot of time with it, but we want to like dive into this. They just got a ton of buzz this week.

Gavin Purcell: And, and, you know, operator is fine. I think that it is fine. It is the first stage of a computer doing things for you, but this supposedly is the next generation, at least a version outside of open AI that is available now to play with. It's impossible to tell what

Kevin Pereira: is hype beast grind SETI anymore on X or threads or whatever else, but people have said, this is, uh, Another deep seek moment for, uh, for open AI.

Kevin Pereira: Oh my word. They're, they're, they're getting their lunch eaten [00:24:00] and we have not tried it yet. So we're going to do it live. Gavin, what task would we want this online researcher to do? It can supposedly research products. It can. Uh, review websites for you. It can do competitive analysis. It can do research on people or brands.

Gavin Purcell: Let's have it do research on Guy Fieri. I mean, we haven't done a good, good, good Guy Fieri bit for a while, but I'd like to know like what it thinks about Guy Fieri from both a culinary standpoint and also like a cultural relevance standpoint, and then, you know, kind of give it the angle of like, We may want to reach out to Guy Fieri for a specific thing.

Gavin Purcell: So it really needs to give us enough information to show us that we are experts in Guy Fieri ness.

Kevin Pereira: Okay, so I'd say, uh, give me research on Guy Fieri, uh, culturally and culinarily speaking. We want to reach out to him and show we're experts. So, um, what, provide, um, Like what? Five, uh, five, uh, in incredible opening lines to say, or something like that.

Kevin Pereira: Like, yeah, five

Gavin Purcell: incredible opening lines to say in one piece [00:25:00] of information that will shock him that we know.

Kevin Pereira: So Gavin, I've run the search and I can see here on the left hand side, just like any other chat, GPT or anthropic, any other traditional LLM, I have my conversation stored there in the middle. I can see that it is doing some research and it's researching the background.

Kevin Pereira: And it's saying, I'll open the Wikipedia page about Guy Fieri to gather detailed information. And then on the right hand side, there is a proxy view panel, um, which it's showing me what the, you know, the web proxy is going through. And I can see that it is in fact, navigating the Wikipedia page. Now it says that it's starting.

Kevin Pereira: And so I don't know how in depth. How long is this going

Gavin Purcell: to take either? Right. I guess we will

Kevin Pereira: just

Gavin Purcell: wait for it to finish.

Kevin Pereira: This is the future. Yeah. I mean, we couldn't come back

Gavin Purcell: to it. That's right. We can come back to it later. Let's let it think for a little while. We'll knock out another quick story that I think is interesting that we should just discuss Mira Moradi.

Gavin Purcell: The former open AI chief technical officer has come out of stealth with her new company. It is called thinking machines, [00:26:00] Kevin, and they have done what all great AI companies do is they released a very long, Essentially blog post on their homepage where you can read about what they're going to do. The big news here is that they really want to focus on the interactions between humans and AI.

Gavin Purcell: So maybe that means more consumer based product. Maybe that means more a little bit about how we interact with them rather than frontier models. But the big thing to know here is that Mira has collected. A lot of really smart people from the AI space, including a lot of former open AI, um, employees who have gone to work with her.

Gavin Purcell: So this is yet another big AI company that's, you know, taken in a fair amount of money. If you missed it, there was a story last week where Ilya Sutskever, another former open AI guy raised, uh, is raising a couple billion dollars at a 30 billion valuation for his not public at all, a super safe intelligence company.

Gavin Purcell: So the open AI alumni are out there doing their, what they do, which is raising a lot of money for something that we're not entirely sure what it is yet.

Kevin Pereira: Hey, you know what, you know what, Gavin, congratulations to [00:27:00] everybody getting billions of dollars in the AI space, because they're going to need it to afford a fleet of dancing robots, which they will all want.

Kevin Pereira: This will be the new flex. It used to be 22 inch rims that spin when the car is stopped. And now it will be a choreographed army of unitry robots. I'm

Gavin Purcell: calling it. Honestly, Unitrix keeps coming out with these videos that are so impressive that every time I post one or see people talking about it, people think they're faked.

Gavin Purcell: And I think, I do want to be very clear about these videos. Like, so many people are like, this, this is fake, it's fake. This is not fake. We have seen multiple years of these robots get better. So, if anybody in your life thinks that these robot videos have been faked, you need to start kind of going back and seeing how they've gotten better.

Kevin Pereira: No, this is the wild thing. So Unitry releases a video, right, of, of their dancing humanoid robot. And it's moving around and people like, this is CG because the leap between what we saw last and this, it's just, it's, it's order of magnitude better. It's so fluid. [00:28:00] It's hip joints are moving. It is, it seems amazing.

Kevin Pereira: It must be CG trick trickery. There must be someone behind the scenes. There has to be something happening. So they released another video basically of the same dance, but now the robot is in front of a mirror, so you can see all around it. Not that it couldn't still be CG with that, but there is that. And then they're throwing things at it and they are thwacking it with sticks and boards while it is dancing to somehow show that it is a real demo.

Kevin Pereira: And I, I mean, you believe it's real now. Yeah.

Gavin Purcell: I mean, I believe these are real for a while. Like I said, I think the interesting thing here is that, like, it's gotten better. We've seen these get better. And also, like, if you'd really need stuff is fake. Go back and look at Boston Dynamic videos from a year or two years ago.

Gavin Purcell: They were already doing stuff that felt really interesting. The thing that's different about the unitry videos that are showing now is that they're taking like we've talked about in the show before simulated, you know, simulated training and then putting it into the robot. So when it looks like it's fluidly dancing, it's actually Yeah, it's not like coming up with this dance on its [00:29:00] own.

Gavin Purcell: There was a dance that it trained on. So it's really good at it. Then the other thing we have to talk about in the robotics world is this insane video from Clone Robotics, which to me is the opposite of what the Unitry video, the industry video, like, oh, this is cool. I feel good. The Clone Robotics video makes me feel like we are down a dark path.

Gavin Purcell: So just to describe this video, if you're listening, everything

Kevin Pereira: about this video is wrong.

Gavin Purcell: EVERYTHING! Clone Robotics is the company that makes those very creepy, like, skin looking robots, first and foremost. So you know that Musculoskeletal

Kevin Pereira: androids, guys. Yes.

Gavin Purcell: So in this instance, they've hung the robot so that its feet are just barely off the floor.

Gavin Purcell: It's dangling! It's dangling and just kind of like Twitching? I just want to know, Clone Robotics, what made you think this was the video to put out? Also, listen to this. If you can hear this, I don't know if you can listen to this. Oh, is that the sound? Okay, sorry. This

Kevin Pereira: is the sound that they chose to play behind the unveiling of the first [00:30:00] bipedal, vascular, skeletal android.

Kevin Pereira: Yeah, it's a Skynet heartbeat as this poor, helpless, uh, Westworld, like, caked in white robot helplessly dangles. It's above the cement, it is creepy, why did they do that, why did they play harvey the wonder hamster

Gavin Purcell: or some happy like Exactly. We are big fans of the Unitry social team and not such big fans of the Clone Robotics social team.

Gavin Purcell: Clone Robotics, come on. Let's do something fun with this device. I already decided to make it creepy if I'm making it. Yeah, yeah. That's pretty good. Will, can we do that? Let's see what it looks like with the Akati Sacks. Maybe it'll be more

Kevin Pereira: entertaining. As it just flails about. Oh, we got an update. We got an update.

Kevin Pereira: Sorry, Gavin. We got an update. We got an update from Convergence. Here we go. It's Finished. It says, here are five compelling opening lines and one shocking fact about Guy FII that you can use in your outreach. Your unique blend of rowdy entertainment and genuine food appreciation has not only entertained, but also [00:31:00] educated millions about the diversity of American cuisine.

Kevin Pereira: Fair.

Gavin Purcell: That's fair. I would agree with that. It's a little bland in terms I wouldn't. Say it that way. I'd be, I wouldn't throw a couple, uh, you know, um, donkey sauces in there at some point.

Kevin Pereira: That's what I'm saying. Okay. So there, uh, oh, your role in raising over 20 million for restaurant workers during the pandemic showcases your impact beyond the kitchen.

Kevin Pereira: True. I mean, that was big.

Gavin Purcell: I mean,

Kevin Pereira: guided a lot during the pandemic. But is that an opening line? No, probably not. No, that's a cool factoid. Unless we're trying to make it more reasonable for us. It's an interesting, yeah, it's an interesting research fact. Okay, shocking fact. Despite your penchant for bold American flavors, your culinary journey started with a love of French cuisine during your high school exchange in Chantilly, France.

Kevin Pereira: Hey, that's pretty good. If that is, now the question is, is it true? Third war, third

Gavin Purcell: war. That's pretty good. What if we could bring out like the The menu that he made from France in high school and be like, guy, what is this? Tell us what this is. And he'd be like, [00:32:00] I don't know how you found that, but it's crazy.

Gavin Purcell: I have made a donkey sauce was actually a little gay sauce in France. The first time that's where it all started. You guys are amazing. So anyway, not bad. That part was pretty good. Again, if it's real, should we, should we double check that? Let's just do a double check.

Kevin Pereira: Like it says during, uh, now this is according to Gemini, but it's basically, pulling from his Wikipedia that Guy studied abroad in France as an exchange student during his junior year of high school.

Kevin Pereira: Again, if you search like, Guy Fieri High School, like, Gemini is surfacing the exact same thing. Yeah, yeah. But Kevin, we didn't have to do

Gavin Purcell: it. We got to talk about other things while it did it for us. That's the key. We just had to go sign

Kevin Pereira: up, give it the task, make sure it was operating, and then go off and do something else.

Kevin Pereira: You're right, Kevin. It's basically the same. So let's get back to these weird Uh, robots that are dangling and dancing, right? Those are the two modes that we've got right now. Dangling and dancing. Um, but soon We'll also unlock destruction, Gavin, because I have to shout out the Sean Ryan Show interview with Palmer.

Kevin Pereira: [00:33:00] Lucky, uh, I listened to all three and some odd hours. Oh, wow. It's that long. It is long. It is winding. I was captivated by way by it. I am a, I'm a Palmer lucky fan boy. Yeah, you are pretty. So Palmer is, um, the, uh, Wunderkid that developed Oculus in his garage and recruited John Carmack and eventually sold, um, Oculus to, to Meta and was unceremoniously ousted, uh, from Meta or Facebook at the time, something he goes in depth on, um, quite a bit, but then he started Anduril Industries, which is a defense company, essentially making smart weapons and AI powered systems.

Kevin Pereira: For our military that will eventually trickle down someday, potentially into civilian devices as well. So this three hour interview, I feel like Stefan from SNL. It has everything. It's got loitering munitions. It's got AI powered Marines. It's got exoskeletons. It has neuro lattice discussions, but.

Kevin Pereira: Palmer's bet [00:34:00] years back was that AI was going to disrupt a lot of things, including defense and the defense

Gavin Purcell: industry, which is a big one, right? Like there's a ton of money. I mean, people understand this, I think a better now, but there's a ton of money made, wasted and spent in the defense industry. And Palmer really said, like, I think I can make this better.

Kevin Pereira: Yeah. And you know, the, the, the pitch, which he, he mentioned several times was we can save, um, Taxpayers hundreds of billions of dollars by our company making tens of billions, and that's a good business and a good value prop. And so, uh, he he's a fascinating, if not for some polarizing figure in the artificial intelligence and tech space in general, but I mean, the interview, I think, is just a phenomenal deep dive into his brain.

Kevin Pereira: You're gonna hear things like exoskeleton suits, which are, um, you know, right now bolted on to Palmer's legs and allowing him to run like sub eight minute miles because they're enhancing his muscle movements. But talking about wiring those and to be clear

Gavin Purcell: on that, Kevin, he's actually using these right now, right?

Gavin Purcell: In the [00:35:00] in the interview, he talks about the idea that like he's not like in that great of shape, but the fact that he put these exoskeleton legs on allowed him to run an eight minute mile. That is feels like future stuff, but it's actually in work there. Yeah,

Kevin Pereira: no, it's they're building prototypes and iterating fast.

Kevin Pereira: And in the future, that could be a device that lets you lift thousands of pounds of weight or run miles without being exhausted. It's just going to amplify your muscle movements.

Gavin Purcell: I know what this is all about. This is like we're moving towards MechWarrior or what is the Kaijus? We're going to giant, giant Kaiju robots.

Kevin Pereira: That's right. Yeah. And, um, you know, he also talks about, um, Um, I think it was IVARS is the acronym, but basically a, a collaboration between Anduril and Microsoft and the Department of Defense where they're building a divisor of the future where what you see, Gavin, can be piped to my headset as well. And I can have, um, uh, access to an entire visual spectrum and I can identify, um, threats because I can even use, uh, [00:36:00] bounced radio frequencies to kind of see through walls and see what's like, there's, I'm doing it a terrible, terrible disservice by trying to grossly distill three hours into a, Oh my God, you got to catch this.

Kevin Pereira: But even discussions on like the loitering munitions, where they've got essentially rockets or missiles that can fly around a target and wait and use visual systems to confirm, Oh, that is the person and, or thing that I need to go and eliminate. They can just kind of go and hang out and then go and target something like that, that's what they're doing.

Kevin Pereira: Go give it a skim and one of the cool things you can do with AI powered missiles is Send them into an area, and then have them do precision targeting, collaborating with each other. Even if you don't have a person there who's, let's say, steering a coded laser target designator. Like, in the past, you would need somebody to say, like, get out there, shine a laser at the thing you want it to go after, and then it goes after that laser dot.

Kevin Pereira: Well, what if instead, you get a missile show up somewhere and say, oh, that's a Li Yang destroyer. I know that the engine room is right there, and I [00:37:00] know that the thinnest part of the deck is right there. If I approach it this angle and this speed from this direction, I can punch through the engine room and destroy engine one.

Kevin Pereira: And if that's successful, another one's going to come in behind me from the other side and destroy engine two. If I miss, or I don't make it through, I'm going to have it instead go for engine one. If you're not getting the video, he's saying all this in his, uh, trademark Hawaiian shirt. It's his patented Hawaiian shirt while holding a gummy bear.

Gavin Purcell: He's got an orange gummy bear in

Kevin Pereira: his hand as he pontificates about future missiles. But like, he's building it. They're doing this stuff.

Gavin Purcell: No, and listen, we've talked about Palmer on the show before, and obviously what he's built and what he's done is pretty incredible. There is, like, the other side of this that is very scary to hear, right?

Gavin Purcell: When you compare it to the idea of, like, What we talk about these chatbots and all this stuff that's going to allow you to do these really interesting stuff. One of the things that's kind of scary about this, obviously, is that a huge part of AI in general is building up AI armies, right? And the longest, scariest part of this is like, that is the story of the Terminator, which we're [00:38:00] not going to get into because hopefully we don't get there.

Gavin Purcell: But this is the thing where like, you know, Anduril is a company that's really based on the idea that he's building up this kind of like large munitions for the Americans to specific, to specifically And one of the things he also said in this, which I thought was really interesting was the idea that in China, you know, he talks about the idea of like what their naval capacity is.

Gavin Purcell: And like, this is somebody who's like thought a lot about what China is like. Yeah. And he said like the shipbuilding capacity of China is 350 X times that of America and that every Chinese ship that's built, even if it's a, a ferry that's used to bring people back and forth is built to military specs.

Gavin Purcell: And so there is this thing where like, you know, we talk about Chinese models and we talk about American models and this kind of fight that's going back and forth. Palmer and other people in the world specifically Palantir and the other defense companies are really thinking about this as a actual fight and that is something that like doesn't sink in enough I think with Americans but they're gearing up for the next version of warfare and it's It's [00:39:00] scary, but also fascinating to watch at the same time.

Gavin Purcell: I'm glad he's on our side, is what I'll

Kevin Pereira: say. No matter what your thoughts are, uh, you know, about AI being used for, you know, for, for defense or even offense, like glad he's on our side. It's a fascinating interview. He also talks about, uh, like, uh, Uh, remote operated versus AI operated, like, bipedal robots, and he, you know, he makes a very good case for, like, these artillery, front line artillery systems that require two or three humans to go and load shells and aim and fire, and those are human lives that are very often lost because they're on the front lines, they're a very juicy target.

Kevin Pereira: for your enemy. And it's like, well, what if that could be one robot that is good at pulling levers, lifting heavy thing, loading it in. It might not be as nimble and capable as a human, but you'd rather have three of those being blown up on the front line. So, I mean, it's like very, very serious, very, very entertaining, especially considering how serious the topic matter is.

Kevin Pereira: I just, I, I went back to it three or four times and just [00:40:00] kept listening and listening.

Gavin Purcell: So maybe in the future we'll just have a big open field where we can send our robots and they can send their robots so that nobody actually has to die. That would be amazing.

Kevin Pereira: How about just like, okay, everybody pile up your money and whoever's money is in the biggest pile, you win and we'll set it on fire.

Kevin Pereira: Both sides have to set their money on fire, but it's like a mystery beast challenge. It's a mystery beast thing. Who's going to build the biggest money pyramid? Who's going to build the biggest money pyramid? And it's like, all right, all

Gavin Purcell: right. We should move on to something that's a little bit more lighthearted, but still really interesting.

Gavin Purcell: Microsoft has just introduced a new Muse game AI model. Now this is in research, but it is very interesting. There was an article in Nature that was just published. Satya Nadella talked about it today. And basically what this is, is their version of a world model that's trained on video games, specifically for making video games.

Gavin Purcell: It's called WAM, WAM model, which stands for world and human action model. And they worked on this with Ninja Theory, which is pretty cool. Like Ninja Theory is [00:41:00] a pretty big game dev that's been around for a while. And what's really cool about this, Kev, is they've got some really interesting videos. This was specifically around one of Ninja Theory's games, but this is again, the idea We saw this in Minecraft.

Gavin Purcell: We saw this Google game. Jen has done it is AI generated video games, meaning that this is not pre programmed, but that you create an AI model that allows you to play within. And what I thought was really interesting about this, Kevin is like, this is Xbox studios working on this, right? So this is actual people who make video games, including Ninja theory.

Gavin Purcell: A lot of people have talked about some version of this as a future of a world building game or, you know, imagine a sandbox game in the future, which generates itself in real time. Again, not like, you know, it's something we have seen a couple of times before, but the fact that it's Microsoft doing it within Xbox feels like a step forward.

Kevin Pereira: Yeah. And this is like, Not only letting the AI apparently like control the, the, the player. So it can learn, but you as a developer, like one of the demos they have is clicking and dragging an asset from [00:42:00] like a, like a generator generative panel into the world and boom, magically the object appears and then the AI or the player can interact with it.

Kevin Pereira: I don't know what is. Like proof of concept or, you know, glossy graphics, but it's very clear that that, uh, you know, from asset creation to physics to gameplay loops, like AI is going to inform all of the things.

Gavin Purcell: Yeah. And I think again, this is Satya Nadella himself tweeting this out along with a very fascinating, uh, quantum computing advancement.

Gavin Purcell: But like, what you can go read about, and you should, we don't have time to get into here, but like, Microsoft is definitely making, oh, what happened? What's going on? Gavin, hot off the presses, hot off

Kevin Pereira: the presses, Gavin. Uh, surprise, surprise, surprise. This is, this is, this is new to you. This wasn't in the rundown, this happened organically.

Kevin Pereira: If I could spend 10, 000 on Amazon products for the host of the AI for Humans podcast, what big purchases would they appreciate the most? Give me highly reviewed products and reasoning. This, Gavin, was the surprise task [00:43:00] that I gave Convergence and their hot off the shelf proxy AI agent. Would you like to hear some of their results after it had spent minutes exploring product recommendations for the AI for

Gavin Purcell: humans

Kevin Pereira: host?

Gavin Purcell: So these are for pretty expensive purchases from Amazon, essentially.

Kevin Pereira: Well, look, we know our audience. We call them the Sugar Tats. We'll put a gift link up for these. We'll put a gift link up for these guys. The Sugar Squad. This is shoutouts to the Sugar Squad on our Patreon, which is real. We have a tip jar, which is a Patreon.

Kevin Pereira: You can give us five dollars a month. But I figured someone out there wants to be a whale, Gavin. Someone wants to spend ten grand on us. Here are some quote, highly reviewed Amazon products that the hosts of the AI for Humans podcast might appreciate. Number one, Instant Pot Duo Plus, coming in at 49. 95.

Kevin Pereira: Okay. This versatile cooking gadget can replace nine common kitchen appliances and is perfect for hosts who enjoy [00:44:00] quick and easy meal prep. Okay. Okay. Starting off strong. Number two. Carbonated bubble clay mask at 8. 65.

Gavin Purcell: Oh my god, oh lord. So basically what it did is it found like highly reviewed products.

Gavin Purcell: Didn't take us into account at all. Not at all. Also not the 10, 000 limit either. No, not even

Kevin Pereira: close. No, the most expensive item is the portable espresso maker at 54 and I can see that it went to, it did basic Google searches again and it says, you know, one of the Buzzfeed articles is 69 things you can get on Amazon that you should just own already.

Gavin Purcell: This is the problem with AI agents in general is that I still think there's a layer and people have talked about this about agents talking to agents to give them better information because the actual web is still garbaged up with stuff like that BuzzFeed article, but Kev, here's another, this is interesting.

Gavin Purcell: This transitions us into this speaking of what, you know, [00:45:00] AI agents early days were supposed to look like humane, the, the AI pin that would wear it in your shirt and pull your shirt down. Cause it was so heavy. 200 million has been sold to HP for 160 million, which is way less than the investment money was raised.

Gavin Purcell: And people are saying mostly for their talent, meaning that they're engineers, this, this pin that was like a big story and kind of almost immediately flopped as a product has now kind of gone dark. In fact, there's the funniest thing I saw in here was They, one of the things they mentioned in the announcement on their website is can I still use my AI pin for offline features because it is no longer connected to the cloud.

Gavin Purcell: So, and their answer is yes, after February 28th, 2025, AI pill, AI pin will still allow for offline features like. Battery level, et cetera. So if you have an AI pin, you paid 700 for an AI pin, you will still be able to check your battery level, which is very exciting.

Kevin Pereira: I wonder if even getting the time of [00:46:00] day is going to work in an offline mode, because it probably needs a cellular modem for that.

Gavin Purcell: I feel like you're right, Kevin. I think the big thing here is like, even they had a T Mobile deal that was set up. The long story short is here is this company was very early. They promised a lot of stuff. This came out around the same time as the rabbit R1, which went through a similar sort of thing that still exists.

Gavin Purcell: And it's still being updated. I think that in general. The AI device conversation just was way too early to come out when it did. Like this was just not ready. There are a lot of AI devices in the works. We know that Sam Altman and Johnny Iver are working on some version of an AI device or a phone that's supposed to replace the iPhone.

Gavin Purcell: So that might come later, but Kev, this is a good chance for us to maybe just do a little pour one out song for, for humane.

Kevin Pereira: Well, yeah, let's, let's take a second, Gavin, along with our audience and just, uh, pour one out. For the humane A. I. pen.[00:47:00]

Gavin Purcell: Say we must pour one out for You made the A. I. in You made bread again You make it for the laughs For the tea You made You burn

Gavin Purcell: for the

Gavin Purcell: All right, everybody. It is time to go through some of the stuff that we love that you made or other people made on the internet using AI this week. It's time for AI. See what you did there. Sometimes you're [00:48:00] scrolling without a care, then suddenly you stop and shout.

Gavin Purcell: Very fast, Kevin. The first thing I want to talk about is a very dumb post from Reddit where this person started making funny yearbook pages, uh, using kids faces. This is just a fun way of showing how AI can still be broken in a fun way. This is a great, uh, you see this guy's image, which is like, All these kids, there's one weird kid in the middle of these giant ears and kind of hairy face.

Gavin Purcell: So I went and did this myself that the prompt the guy used was funny yearbook page with odd children names. It doesn't do amazingly, but you do get some funny results. So anyway, this is just a fun AI prompt to try with your favorite image model.

Kevin Pereira: I mean, this is if someone made a bizarre, like AI guess who game?

Kevin Pereira: Yes, exactly. Because it's way more innovative, but I love that in between like seemingly normal, but Kind of [00:49:00] bizarre children there are just horrible monsters like creepy alien gremlins like and and weird ones with mustaches like it's delightful and and your swing at it is great too it's very fun not necessarily a reddit post but instead uh like a a shocking scientific leap potentially there is a thought to action model gavin um that uses uh megs or eegs basically reading uh your brain activity uh They have a model that can predict with like 80 percent accuracy in near real time what it is you're about to go type.

Kevin Pereira: So you think about typing the thing and out comes the text. And this could be a brave new way that when we talk about, you know, Extended reality or augmented reality, having glasses on that are displaying the future rather than having to speak things into existence. If you're in a place where you can't, you might be able to just think the text input that you want to see on the [00:50:00] screen and there it appears.

Gavin Purcell: Yeah. So this is, to be clear, this is meta publishing and research papers. And, you know, meta has like. So many crazy scientists working there. And sometimes this stuff doesn't get enough shout out, but their, their paper is really interesting. Cause basically this is, you know, Elon's been working on the, uh, the brain interface, uh, neural link for a while, but what they're trying to say here is that you may not need to have an implant in your brain, but that you could actually use an EEG device, something that can read your head.

Gavin Purcell: Or maybe a giant thing on your head. Like I think maybe in the future, one of these giant things on our heads, like electronic Afros that will allow us to read our brains. The funniest thing to me, Kevin, is in this video and in the in the paper, they have this person sitting in what looks like just this giant mind reading chair, which is like, Which is like the weirdest looking device, but it kind of is like what Professor X would use, like to kind of read the world at large.

Gavin Purcell: So maybe this is the world we're getting into. Maybe we're going to get into just giant mind reading devices and we'll all be able to read each other's minds all the time.

Kevin Pereira: Put me in my digital cocoon and let me think. [00:51:00] All of my spicy chats. Deep thinking? You're big brain? You

Gavin Purcell: want big brain? That's right.

Gavin Purcell: That's right. Okay, so keep going here. There was a very cool post from somebody named Sanimatic. And Sanimatic is an ex user. And this was just a good example of showing The kind of AI, uh, art haters out there, what a real artist can do in a short period of time using AI tools. And now I don't know exactly what tool this guy was using.

Gavin Purcell: It looks a lot like what Korea does as a, as an image to image model. But basically this was an artist who made like a very kind of rough, animatic looking design of two people in a spaceship. Didn't put a crazy amount of details and then put it through a, an image model and then animated it with a video model.

Gavin Purcell: And you just. See the quality of what this person made to start with, because it's clear this person is an artist immediately transformed into a scene from a film. And I just thought, Kev, this is like the perfect thing to show people who are like, Oh, AI is not art or AI isn't useful for artists. Like this [00:52:00] is the exact pipeline of how AI is useful artists and what it will do to the Hollywood pipeline in general.

Kevin Pereira: Yeah, it's an, you please check out the show notes if you're getting the audio only and see it. It's a quick little video snippet, but it it's really inspiring for that exact pipeline that Gavin is talking about. And I think, uh, similarly, Cliff Bleszinski, uh, historic game designer from friend of ours.

Kevin Pereira: He's the guy we know personally. Good, good guy. Yeah. Yeah. Cliff is great. And you know, everything from Jasmine jazz, jackrabbit to unreal tournament to of course, gears of war, uh, fortnight, even he had his hands in all of the things. He's taken quite a break from game design, but he's been dipping a toe back in the waters and he shared a Dropbox, uh, with his, uh, on his Twitter for a game that he had created called Panic Garden.

Kevin Pereira: And what's, what's great is like, if you have any interest whatsoever in game or game design, it's fascinating to go through and see the way the documents are written up, the way he describes the world, the mechanics, the characters. But he also shares mid journey folders where [00:53:00] he is using AI art to, to develop proof of concepts.

Kevin Pereira: And it's fascinating, Gavin, because. I mean, A, it's just great, great peek inside of the creative process on that side of things, but also because like the, the gaming community specifically has been one of those very vocal opponents to AI touching anything. And yet here, I think because it is such a notable Game designer saying, look, I'm using these tools to try to get my vision across.

Kevin Pereira: People seem to be a little bit more accepting of the fact that there's AI involved in this pipeline. And honestly, I think they should be.

Gavin Purcell: And honestly, conceptual art is a great use case for AI. Now, some people are always going to have the hate on AI and they're not going to use it at all, but like, it does allow you to make these generations very fast.

Gavin Purcell: And to talk about Santa Maddox, uh, work that we just saw. Even that would be a really interesting use case for an artist, right? So that you could take this kind of like basic shape or kind of look of what you wanted and then make this screenshots from that, you know, uh, Cliff is using mid journey straightforward, but like there's [00:54:00] all sorts of cool AI tools to be able to do.

Gavin Purcell: So that was very awesome. Finally, we have what I think is a really fun. Fun real world use case of robotics. This is a video that features Stephen Curry from the Golden State Warriors actually using a robot, a robot to return basketballs to him. And if you want, if you're not watching the video, what's interesting about this is the robot kind of reminds me of those robots that Amazon has designed for their warehouse.

Gavin Purcell: It's not humanoid. It's literally a basket that moves around on the floor to find where the ball is going to go and then spits the next ball out at him. This is the kind of functional robotics that I think are really useful in the real world. And we're going to see a lot more of.

Kevin Pereira: It's such a weird looking little robot, but it makes sense.

Kevin Pereira: And in the background, I don't know if you noticed some Boston Dynamics, uh, like robot walking dogs. There's some out of focus dogs prowling around in this Oracle branded gymnasium. There's, there's a lot going on in the video, but it's clear. It's a, it's interesting use case of robotics, but like the moment I saw it, I was like, Toyota is saying, hold my beer.[00:55:00]

Kevin Pereira: Because if you look at, uh, if you look at the Guinness world records account, they tweeted out, uh, a while ago, this was back in December, but Toyota made a basketball shooting robot that I've been fascinated with since the early prototypes, basketball sinks, the world's largest free throw. So while we've got a little Fisher price, It looks like model after like connect five colors moving about to return you a ball.

Kevin Pereira: Toyota's got the full jump shot happening. Like Toyota's on fire.

Gavin Purcell: Each of these is a step closer to real life cyber ball, which I cannot wait for. I've said that a bajillion times, but I cannot wait for the robot sports to start. Kevin, speaking of robot sports and other great sports. would be to follow our newsletter.

Gavin Purcell: You should definitely spend some time looking at our newsletter. We have a newsletter that comes out every Tuesday morning where we focus on whatever the biggest thing that day is, or some, some creative tools. We dive into it. Please subscribe. It's a very easy thing to go find. You can find the link at AI for humans dot show our website.

Gavin Purcell: It grows as much as our other stuff is growing, which is very cool. It means people are liking it. So go check [00:56:00] it out. So Kev, when it comes to what we did with AI this week, I spent some time. Using still my favorite AI video tool, uh, VO two, which is Google's AI video tool, but they have now released it for the public in the YouTube shorts, uh, uh, generation app, which is actually in your YouTube app.

Gavin Purcell: So in YouTube? Mm-hmm. You can now, if you're in the YouTube mobile app, you can now develop these videos in there. It's not the greatest example of how to work with AI video, but you just basically go into the app. You, you select instead of selecting one of your, sorry, let me pick this up. So basically you go into the YouTube app and you push the plus button as if you're going to shoot an original short in there, which you can do.

Gavin Purcell: And instead of selecting, um, one of the videos you already have on your phone, you push the empty button and what then comes up is a interface. for a video interface. So you can then enter into any, you know, create a prompt for any video. And it's using VO two, a turbo model, not the full model that we've had access to, that's [00:57:00] not out in the world, but a turbo model to create these videos.

Gavin Purcell: And I made this little small video, which was like a storytelling video of like what it would look like if a Terminator robot, our kind of mascot got discovered for a movie and then got put into a movie. And that was fun. It was, well, I should say this. It wasn't exactly fun because The, the VO model is fine, but working within the YouTube app to edit anything is an absolute nightmare and it really made it difficult to do.

Gavin Purcell: I hope they increase the ability for people to use this, but right now, if you're out there, open your YouTube app, you can go in and you can play with VO2 or a version of their VO2 model. And it is very capable. It will be frustrating what it will and won't let you discover or create because there are many times like I wasn't able to generate something with a human in it.

Gavin Purcell: There were other times where it wouldn't return something. You'll see it in this video that I made. There's a couple still shots of people on stage. That's because I was able to get a still image of two people on stage [00:58:00] and in a work show, but I was not able to generate a video of that for some reason.

Gavin Purcell: So. There's all these weird restrictions on it, but it is a cutting edge AI model that is now available to the public.

Kevin Pereira: Yeah, I mean, it's interesting that the hurdles that you're talking about having to leap are not. Issue with the AI model itself. It's a user interface problem. So I'm sure that at some point they will iron that out or pay to win.

Kevin Pereira: But, um, yeah, people should go and try

Gavin Purcell: it out. Right. It's free. It's free. I mean, you can use it if you have a YouTube account, anybody can use it right now, and I think ostensibly they're trying to get people to use this, to kind of spread, you know, make YouTube shorts stronger and do stuff in YouTube shorts.

Gavin Purcell: It's not like you really only get like a five second output and like, you can't really control. You basically are first asked to generate the image. So it's very much like VO2 where it's image to video. You have to first make an image and then you actually make that animated. It's not the greatest control in the world, but it is free and you can do it right now.

Gavin Purcell: So Gavin, what are our key takeaways

Kevin Pereira: for the week? As we look at. [00:59:00] This is our Springer esque final thought. What do we leave the dear viewer or listener with, other than like, subscribe, and comment in Five Star Review? Ooh, I don't know, that's a good question, Kevin.

Gavin Purcell: Okay,

Kevin Pereira: bye!

Gavin Purcell: Goodbye everybody, we'll see y'all next time.