Latest / Elon Musk Thinking / Conversation with Elon Musk and Satya Nadella at Microsoft Build 2025.

Transcript

Conversation with Elon Musk and Satya Nadella at Microsoft Build 2025. · Elon Musk Thinking

• 0:00
  Speaker 1: Thank you so much, Lon for being here at build.
• 0:03
  Speaker 1: I know you started off as an intern at Microsoft,
• 0:06
  Speaker 1: you were a Windows developer, and of course you're a
• 0:08
  Speaker 1: big PC gamer. Still, you want to just talk about
• 0:11
  Speaker 1: you in your early days with Windows and the kinds
• 0:15
  Speaker 1: of things you build.
• 0:17
  Speaker 2: Yeah, well, actually started before Windows with us. I had
• 0:20
  Speaker 2: one of the early IBM PCs with MS DOS and
• 0:26
  Speaker 2: I think I had like a one hundred and twenty eight
• 0:28
  Speaker 2: K in the beginning, and then at double T two
• 0:30
  Speaker 2: fifty six K, which felt like a liar, so I yeah,
• 0:34
  Speaker 2: pro programmed video games indus and then in Windows remember
• 0:41
  Speaker 2: Windows three point one.
• 0:43
  Speaker 1: Yeah, no, it's wonderful. I mean even at the last
• 0:46
  Speaker 1: time I chatted with you, you were talking all about everything,
• 0:49
  Speaker 1: the intricacies of active directory, and so it's fantastic to
• 0:53
  Speaker 1: have you at our developer conference. Obviously, the exciting thing
• 0:56
  Speaker 1: for us is to be able to launch grock on Azure.
• 1:00
  Speaker 1: I know you have a deep vision for what AI
• 1:03
  Speaker 1: needs to be, and that's what got you to get
• 1:05
  Speaker 1: this billed. It's a family of models that are both
• 1:08
  Speaker 1: response and reasoning models, and you have a very exciting roadmap.
• 1:11
  Speaker 1: You want to just tell us a little bit about
• 1:13
  Speaker 1: sort of your vision the capability you're pushing on both
• 1:16
  Speaker 1: capability and efficiency, So maybe you can just talk about
• 1:19
  Speaker 1: a little bit of that.
• 1:20
  Speaker 2: Sure. So yeah, with with GROC, especially with Rock three
• 1:23
  Speaker 2: point five that is about to be released, it's it's
• 1:27
  Speaker 2: trying to reason from first principles, so apply kind of
• 1:30
  Speaker 2: the the tools of physics to thinking. So if you're
• 1:35
  Speaker 2: trying to get to fundamental truths, you you try you
• 1:39
  Speaker 2: boil things down to the axiomatic elements that are most
• 1:41
  Speaker 2: likely to be correct, and then you reason up from there,
• 1:44
  Speaker 2: and then you can test your conclusions against those axiomatic elements.
• 1:47
  Speaker 2: And you know, in physics, if if you violate conservation
• 1:50
  Speaker 2: of energy or momentum, then you're either going to get
• 1:53
  Speaker 2: a Nobel price or you're you're wrong, and you're certainly
• 1:56
  Speaker 2: wrong basically. So so the the that's really the focus
• 2:01
  Speaker 2: of Rock three point five is uh, sort of I
• 2:05
  Speaker 2: find a fundamental physics and applying physics tools across all
• 2:10
  Speaker 2: lines of reasoning and to aspire to truth with minimal error.
• 2:16
  Speaker 2: Like there's always gonna be some mistakes that are made,
• 2:19
  Speaker 2: but we aim to get to truth with acknowledged error,
• 2:25
  Speaker 2: but minimize that error over time, and I think that's
• 2:29
  Speaker 2: actually extremely important for AI safety. So I've put a
• 2:33
  Speaker 2: lot for a long time about AI safety and wealth.
• 2:36
  Speaker 2: Book conclusion is the old maximum that honesty is the
• 2:39
  Speaker 2: best policy. It really really is for safety. But any
• 2:45
  Speaker 2: want to have size. You know, we we haven't will
• 2:48
  Speaker 2: make mistakes, but we aspire to correct them very quickly,
• 2:51
  Speaker 2: and we are very much looking forward to feedback from
• 2:53
  Speaker 2: the developer community to say like, what do you need?
• 2:56
  Speaker 2: Where are we wrong? Cocker, we make it better, and
• 2:59
  Speaker 2: to have Rock be something that the developer community community
• 3:04
  Speaker 2: is very excited to use and where they can feel
• 3:06
  Speaker 2: that their feedback is being heard and GROCK is improving
• 3:12
  Speaker 2: and serving their need.
• 3:13
  Speaker 1: Yeah, I know it's in some sense, you know, cracking
• 3:16
  Speaker 1: the physics of intelligence is perhaps the real goal for
• 3:21
  Speaker 1: us to be able to use AI at scale, and
• 3:24
  Speaker 1: so it's so good to you know, take that first
• 3:26
  Speaker 1: principle's approach that you and your team are taking. And
• 3:29
  Speaker 1: also you're deploying this. I mean one of the things
• 3:31
  Speaker 1: about sort of what you do is you're doing you know,
• 3:34
  Speaker 1: unsupervised FSD on one side, you're doing robotics. Of course
• 3:38
  Speaker 1: there's Rock. You're deploying Rock across all of your businesses,
• 3:41
  Speaker 1: from SpaceX to Tesla. Obviously at X, I would love
• 3:45
  Speaker 1: to even you know, one of the themes for this
• 3:47
  Speaker 1: developer conference, Elan is we're building pretty sophisticated AI apps. Right,
• 3:52
  Speaker 1: It's not even about any one model. It's about orchestrating
• 3:55
  Speaker 1: multiple models, multiple agents, just anything that you're seeing in
• 4:00
  Speaker 1: the real world application side, even inside of your own companies.
• 4:04
  Speaker 1: When you think about even a Tesla or a space
• 4:06
  Speaker 1: X where you put grock and needs the other AI
• 4:09
  Speaker 1: models you're building.
• 4:10
  Speaker 2: Yeah, it's incredibly important for a model to be grounded
• 4:14
  Speaker 2: in reality reality. You know, I was saying, which is
• 4:19
  Speaker 2: like like physics is the low and everything else is
• 4:20
  Speaker 2: the recommendation. Which is I'm not suggesting people break but
• 4:23
  Speaker 2: the laws made by you know, humans. Uh, you know,
• 4:27
  Speaker 2: we should generally avail laws of humans. But but I've
• 4:31
  Speaker 2: seen many people break human made laws, but I have
• 4:34
  Speaker 2: not seen anyone break the loads of physics. So for
• 4:38
  Speaker 2: for any given AI grounding it against reality and reality.
• 4:43
  Speaker 2: For example, as you mentioned with with the car, it
• 4:45
  Speaker 2: needs to drive safely and correctly. Uh, the human raid
• 4:49
  Speaker 2: robot optimist needs to perform the task that that that
• 4:53
  Speaker 2: it's being asked to perform these These are things that
• 4:57
  Speaker 2: are very very helpful for issuing that the model is
• 5:01
  Speaker 2: crucial and accurate because it has to adhere to the
• 5:04
  Speaker 2: loads of physics. So so I think that's actually maybe
• 5:08
  Speaker 2: so some somewhat overlooked or at least not talked about
• 5:11
  Speaker 2: it enough. Is that to really be intelligent, it's it's
• 5:15
  Speaker 2: got to make predictions that are in line with reality.
• 5:18
  Speaker 2: In other words, physics. That's that's it's a really fund
• 5:21
  Speaker 2: metal thing and and being able to ground that with
• 5:25
  Speaker 2: the cars and robots is very important. We are seeing
• 5:29
  Speaker 2: Grock be very helpful in things like customer service, and
• 5:34
  Speaker 2: you know that the AI is infinitely patient and apparently
• 5:37
  Speaker 2: and you can yell at it and it's still going
• 5:38
  Speaker 2: to be very nice. Uh So that's good. Yeah, And
• 5:43
  Speaker 2: so so I think in terms of improving the quality
• 5:47
  Speaker 2: of customer service and sort of issue resolution, AI's were
• 5:52
  Speaker 2: already uh Grock is already doing quite a good job
• 5:55
  Speaker 2: that at SpaceX and Tesla, and and we look forward
• 5:59
  Speaker 2: to like offering that to other companies.
• 6:02
  Speaker 1: No, that's fantastic, Really thrilled to get this journey started,
• 6:06
  Speaker 1: getting that developer feedback and then looking forward to even
• 6:09
  Speaker 1: how they deployed. There is these language models, there's you know,
• 6:13
  Speaker 1: I think over time we will have this coming together
• 6:15
  Speaker 1: of language models with vision with action, but to your point,
• 6:19
  Speaker 1: being really grounded on a real world model, and that
• 6:22
  Speaker 1: I think is ultimately the goal here. And so thank
• 6:25
  Speaker 1: you so much Eland for briefly joining us today. And
• 6:27
  Speaker 1: we're really excited about working with you and getting this
• 6:30
  Speaker 1: into the developers' hands.
• 6:31
  Speaker 2: Thank you, thank you very much. And I can't empsize
• 6:33
  Speaker 2: enough that we're looking for feedback from you that develop
• 6:36
  Speaker 2: an audience. Tell us what you want and we'll make
• 6:39
  Speaker 2: it happen. Thank you,