General Intuition raises $220M to train robot brains on 3 billion gaming videos per year

Speaker 2: Yeah. Well, thank you so much. Great joining, Tom. Have a great weekend, and we'll talk to you soon. Goodbye. Let me tell you about CrowdStrike. Your business is AI. Their business is securing it. CrowdStrike secures AI and stops breaches. We've been keeping him waiting no longer. We got him to win from genuine... Intuition, cofounder

Speaker 1: and CEO.

Speaker 2: Tim, how are you doing? Welcome.

Speaker 1: Man, the bet.

Speaker 5: He's back. Doing well. Thanks, guys.

Speaker 2: Give us the news. You raised some money. How much did you raise?

Speaker 5: You raised 220,000,000 at a 6,000,000.

Speaker 2: Fantastic. Let's get the replay going. Let's get that replay going. Let's see that Gong hit. Thank you.

Speaker 1: With authority, powerful hit. So... For powerful raise. What unlock this new capital?

Speaker 5: Successfully were able to prove transfer to real world robotics from a lot of the game's data. And so I think a lot of the evidence is pointing that you can actually, in fact, just scale pixel space. Humans tele... Teleoperate robots with the exact same brain that they can play video games, and a lot of... And likewise, other way around. Right? People play video games, and then they take those skills, and they they can use the same game controller to teleoperate a robot. And I think a lot of our research, and now we've we've shown some demos as well, is indicating that this transfer works and you can scale pixel to action space on our massive data modes. And, yeah, I'm excited to show the world to models in the next few

Speaker 1: Very cool. Talk about talk about the data mode. I know we... I know we've talked about it probably a couple times now, but for anyone tuning in for the first time, talk about Yeah. How and why you guys are in a position to do this better than most.

Speaker 5: Yeah. So we built the largest we built the largest video game recorder in the world, and so we get a lot of data on how people control embodiments inside video games. So when you're playing a video game, you're essentially just transferring your perception to the screen and then using the controls to control whatever character you have on the screen many times in first person view, just like when you're on a robot. And so because robots also ship with game controllers, we can map those actions, and then largely it becomes just an environment transfer problem where the pixels are different, but the inputs are the same. And so we can pre... Just like transformers or LLMs are pre trained on predicting the next text token from the Internet, we can pre train on the 3,000,000,000 videos that people capture and upload to metal every year, the game recorder Yeah. To go... So to go from frames to actions. And so LLMs go text to text. We go frames to actions, and then we can scale that into extremely general models that can control every robot that ships with the game controller, which is the majority of robots right now.

Speaker 1: Give us your overview of of various robotics categories. Where are you most excited about in the short term? Obviously, in the in in the more... Let's say in the next year, it's all about just working with these companies around like testing and proving out different use cases. But what categories are you most bullish on? Because I think for general intuition, you're... I think it's probably gonna be quite important to hit, like, some of these big inflection points when some of these use cases really start taking off in the way that we've seen other players and infrastructure providers hit, like, the inflection point with, like, personal agents, for example, or or browser automation or these other use cases?

Speaker 5: Our bet is that that inflection point actually sits largely in simulation first, and then most of companies betting on physical world robotics only are going to miss an inflection point because every robotics lab goes every robotics manufacturer, every company goes through sort of testing their robots in simulation before they transfer them over to the real world. And so this year, we've been focusing a lot on scaling with customers in simulation. Next year, a lot of it will be on transferring over to the physical world. Where we are focused specifically at the moment is quadrupeds, drones, and I like stable platforms. Right? Humanoids are are quite unstable still. Their supply chains haven't ramped quite as much, and so... But you can control them using a game controller. And so if your robot ships with a game controller and has a front facing camera, then we can likely work well together, and the models will transfer extremely well.

Speaker 1: Very cool.

Speaker 2: God mode. In video games God mode. It gives you unlimited health. But recently, AI researchers have been talking about a different kind of God mode. Imbuing AI models with God like meaning. What is your take on the the models that you're creating? Will they be moral agents? Will they be have consciousness? How have you grappled with the question of is this god in a box or is it a machine intelligence?

Speaker 5: It's a machine intelligence. I see no reason to believe otherwise. I think our our processes in the brain, our chemical processes, there there's absolutely no reason at the moment to believe these machines are conscious. I think the entire debate is honestly somewhat silly. Mhmm. I... Right? A lot more things would have to be proven to draw any equivalence there. And I also think that, for us, I think the goal is just studying these systems, understanding what they do well and not well, and and and then and then going from there. I think, you know, I try to stay away from a lot of that cultish behavior and just stay grounded.

Speaker 1: Yeah. Talk about sentiment, AI sentiment among among gamers because anytime we've had somebody in the the gaming world, sort of legacy gaming world on the show and you bring up AI, they go, oh, you know, change the topic, you know, no one no one wants to be caught saying anything positive about it. But over the last couple weeks, I feel like people are suddenly creating a bunch of crazy experiences, even home hackers porting games from here here to there. And and and it feels like we could be at at a moment where people are like, hey, maybe this thing is actually pretty cool and a useful tool.

Speaker 2: You could be looking at the future Ben Affleck of gaming.

Speaker 1: Oh, that's right. Potentially.

Speaker 5: I thought that clip of Ben Affleck explaining freezing weights and and and it was so incredible. I I I... That made my day, honestly, seeing him explain that. I... Okay. So on gamers, why why are gamers frustrated with AI or historically have been? Because it really hadn't done much good for them. Right? It made their GPU prices go up. Sure. A lot of cap... Right? There was a bunch of games that were delayed from studios, and a lot of their investment and a lot of these large companies was going towards AI. Meanwhile, all the developers were being laid off at game studios, so it's obvious why gamers were frustrated. And I think the reason was it just hadn't done much for them. And I think now we're starting to see the early signs of that changing. So modding, obviously, becoming a ton easier. I think in our case, we'll see different types of games being possible on top of these models and different types of interaction patterns. And it just... Like, AI just hadn't done anything for them. And I think we're now turning the corner. I I do think it's starting to show, and we just need to continue creating better experiences for gamers and and bring them along in the journey as opposed to the other way around.

Speaker 2: Very cool.

Speaker 1: Awesome. Well said. Congratulations. An amazing progress.

Speaker 2: Yeah. Really amazing progress.

Speaker 1: I love our our our favorite part of doing these interviews of seeing snapshots of companies as they as they make process progress between rounds.

Speaker 2: I I I have I have one more question about Yeah. You know, like the the... It's not necessarily the anthropomorphization but just this idea that like some people like just hanging out and chatting with the LLMs And you have metal and people are gaming and uploading clips. And I'm wondering if you think that there's a... That there will be a real market for a... An AI companion that you play video games with. Like you queue into Fortnite, you want your squad to be full and so you populate that with something that's roughly at the same level as you but can also chat and do callouts and has a certain flavor and and texture and humor to it. Do you think that that will be something that we'll see? Or is the structure of the video game industry such that that will just get banned?

Speaker 5: Yeah. I I think for sure we'll see this because when you're launching video games, your largest problem is use... Usually player liquidity Sure. Which largely drives for attention.

Speaker 2: Yep. And so I

Speaker 5: think if you're a game developer, you want these type of experiences, and I think you want your environments to be really interactive and rich. Yeah.

Speaker 2: And so

Speaker 5: I think naturally, the economics of the industry will push towards this being a thing. I think there's a debate of... To the extent we... You should label that. Right? You should let people know that they're interacting with AI systems. But I do think it will be a thing, and I think, you know, there will be lots of good things that come from it, there will be bad things that come from it. And we just have to study those and evaluate them as they come.

Speaker 2: Yeah. But it'd just be like a really fun NPC as opposed to just like the generic bot there and counterstrike.

Speaker 5: Exactly. Exactly. I think a lot of it will start... You know, the best things start out looking like a toy, and I think the same thing will be true here.