Figure ended its OpenAI robotics partnership to build in-house

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

1. 1:45:57

   You had a pretty close relationship with OpenAI, correct? Yeah, they led my, so Sam and OpenAI led my Series B, co-led my Series B with Microsoft. That was a few years ago now. So we raised about a little under 700 million in our Series B, our second round of funding. And they joined my board and then we ended up spending basically a year with them working on, well, I mean. I'll give you the background.

2. 1:46:27

   The goal was to try to advance AI models for humanoid robots together. And they have some great folks that have worked on LLMs and chatbots and things. And at the time, we had a full AI team internally. So we were basically working weekly, daily, on basically how do we advance state-of-the-art language models for robotics. And, you know, like, yeah, I ended up firing them a year, a year later, but in splitting ways, but listen, they're a great team.

3. 1:47:09

   I like the senior leadership and everybody there, Sam included, like we're great to interact with. um the issue lied for us of like um there's nobody's ever put like advanced like language models into these systems and made it we have to like produce like action output on the robot and it's like a very different thing than like um next token prediction for like language models um we ended up finding that the team we had in place you know my team lead the folks we have here all from google deep mind or certain areas of like you know top ai programs and they're really good. The team now we have is over 50 or so on our Helix team internally. We just found that like that team we had internally, we just ran like kind of circles around them, like every day. We had a hard time getting, like, you know, in robotics, you gotta like run the robot, see how it does.

4. 1:48:00

   You know, like you want to run a new like AI experiment or do some ablations, like and some evals, you need to like run the robot at the end of the day and see how it does. Like sim is one thing, you can get certain far running simulations and looking at loss curves and stuff, But at the end of the day, we need to see how the robot does. And we just had a hard time getting him in the office. We had a hard time basically advancing stuff together as a team.

5. 1:48:24

   Ended up, the strategy we had internally and the team we had was just complete superstars. They're the best robot learning folks on the planet that set a figure. And it got to a point where I got a call one day. We were also week to week showing them how we were doing all this work. And I got a call one day saying like, hey, we've been watching your progress, it's unbelievable. And we're thinking about doing robotics work internally.

6. 1:48:55

   And I was just like, this is over. Like, yeah, just get out of here. Like, this is like, we're teaching you how to do like robot learning. You're seeing our progress. We had like a couple of the, you know, Sam and a couple of the co-founders on site at one point, we were right before this and they saw it and they were like, wow, this is like, it was doing like this neural network on the table and they were just like, Jesus, this is amazing.

7. 1:49:15

   Shit. And I was like, you know, they were still at a point where they continue to want to work together after this. And I was like, there's no way we're gonna teach you how to do this stuff anymore. And also, we just like got no value out of the whole relationship, very little. I mean, listen, it was helpful having them lead the round. It was like, there was some good brand association there, but like beyond like that, there wasn't much.

8. 1:49:37

   So we ended up, you know, we're gonna chart our own territory. we're going to do AI ourselves here. It was also just became like a, to be frank, like it became like really hard to recruit. We were like, you know, I, I have to spend a lot of my time hiring like, like on the AI team and, um, we bring candidates in and they'd be like, oh, you guys do the robot opening ideas and models. And I'm like, oh, no, not really. No, we have a whole AI team internally. We do model development here ourselves. Uh, you know, like we're advancing ourselves and it just wasn't the perception from the outside. It was just hard. Uh, so that also wasn't helpful for us both hiring was not great and we were like you know there was like an information passing back that I think wasn't really helpful for us long term if we're gonna be competitors so we decided to split ways

9. 1:50:19

   I decided specifically to split ways but they have a great team I think they're doing robotics now internally Um, sounds like it. Yeah, exactly. Yeah, yeah, exactly. Uh, I was like, I got a call saying