No Priors: Artificial Intelligence | Technology | Startups

Bob McGrew argues superintelligence requires physical infrastructure and flexible manufacturing, not just digital compute

60 min Listen to the original
Artificial IntelligenceRoboticsManufacturingDefense TechnologySpace ExplorationFermi ParadoxEnterprise Software

Summary

Bob McGrew and other speakers argue that superintelligence is constrained by physical resources like GPUs and wet labs rather than digital compute or researcher automation. McGrew explains that Arda's mission is to solve American manufacturing bottlenecks through flexible automation, partnering with companies like Anduril to reduce build times. The discussion contrasts OpenAI's technology-first scaling approach with Palantir's problem-first forward-deployed engineering model. Speakers also explore the robotics ecosystem, predicting a tension between generalist models and specialized fine-tuned solutions. Finally, they discuss the Fermi paradox, suggesting that self-replicating machines expanding at light speed explain the lack of observed alien civilizations.

Episode sections

  1. 0:00 – 2:55

    Introduction and the Physical World Bottleneck

    Bob McGrew argues that superintelligence requires changes in the physical world, noting that digital AI advancements have not yet significantly altered physical reality. He uses the example of Waymo self-driving cars, which remain constrained by physical traffic, to illustrate that digital intelligence alone does not solve physical limitations.

  2. 2:55 – 6:25

    The Hard Takeoff Argument and Compute Bottlenecks

    McGrew rejects the 'hard takeoff' theory, which posits that automating researchers will lead to runaway self-improvement. He asserts that researchers are not the bottleneck; rather, physical resources like GPUs, compute capacity, and wet labs are the limiting factors. Automating researchers provides a one-time efficiency boost rather than an infinite acceleration loop.

  3. 6:25 – 13:22

    Labs vs. Startups and the Scarcity of Physical Assets

    The speakers discuss whether labs or startups will drive progress in robotics and biology. McGrew suggests that as intelligence becomes cheap, physical assets like wet labs and manufacturing capacity become scarce. He implies that frontier labs may not have a comparative advantage in building physical infrastructure compared to specialized startups.

  4. 13:22 – 16:33

    Arda's Mission and American Manufacturing Challenges

    McGrew explains that Arda aims to make American manufacturing competitive by addressing high labor costs and the inflexibility of current automation. He notes that while the US excels at design, it struggles with manufacturing due to the high cost of retooling automated lines for frequently changing products like consumer electronics.

  5. 16:33 – 18:42

    Flexible Automation and Defense Applications

    Arda focuses on industries requiring high mix and speed, such as defense, consumer electronics, and data centers. McGrew highlights the defense sector, where rapid design changes are necessary due to battlefield dynamics, as a key use case for their flexible automation technology.

  6. 18:42 – 25:53

    Robotics Form Factors and Closed-World Problems

    McGrew distinguishes between open-world robotics (e.g., cleaning a bedroom) and closed-world robotics (e.g., assembly lines). He argues that assembly lines are closed-world problems where traditional automation can work if made flexible, and that the primary barrier is not the capability of robot arms but the economics and speed of reconfiguration.

  7. 25:53 – 28:40

    Arda's Partnership with Anduril and Product Development

    Arda partnered with Anduril to build a software platform for manufacturing engineers, aiming to create a 'Cursor for manufacturing.' They have deployed this across Anduril's product divisions, including drones, missiles, and the Fury combat aircraft, reducing virtual build times from days to minutes.

  8. 28:40 – 30:05

    Forward-Deployed Engineering and Product Strategy

    Speaker_0 explains how Arda uses design-for-manufacturing feedback loops to optimize production costs and automate stations. Speaker_2 notes this pragmatic approach leverages off-the-shelf AI rather than building massive proprietary models, contrasting with earlier hardware-software attempts like 'Git for hardware.'

  9. 30:05 – 34:13

    OpenAI vs. Palantir: Technology-First vs. Problem-First Approaches

    Speaker_0 contrasts Palantir's problem-first, forward-deployed engineering model with OpenAI's technology-first approach. He recounts a disagreement with Ilya Sutskever, who argued that scaling training data on general internet data was superior to learning from individual customer problems, a stance Speaker_0 disputes for practical application.

  10. 34:13 – 41:07

    Robotics History and the 'Bitter Lesson' in Physical AI

    Speaker_0 details his journey from skepticism about neural networks to working on robotics at OpenAI, including a project to solve a Rubik's Cube with a Shadow Hand. He argues that recent robotics breakthroughs stem from large models predicting physical data patterns learned from internet videos, rather than massive proprietary robot datasets.

  11. 41:07 – 45:02

    Robotics Ecosystem: Generalists vs. Fine-Tuned Specialists

    Speaker_0 predicts a tension in the robotics market between generalist models like Figure (Fable) and specialized fine-tuned models. He suggests generalists are slow and hard to fine-tune, creating an opportunity for companies like Pi or General Robotics to offer faster, use-case-specific solutions, while edge devices will require distilled, open-source models for safety and speed.

  12. 45:02 – 51:37

    Scaling Enterprise Adoption via Forward-Deployed Engineering

    Speaker_0 describes Arda's use of the forward-deployed engineering (FDE) model, originally developed at Palantir, to embed engineers with customers like Anduril. He emphasizes finding 'power users' to iterate quickly on products, contrasting this with traditional enterprise sales, and stresses that FDE is a product strategy to build leverage, not just consulting.

  13. 51:37 – 56:17

    The Physical Singularity and Self-Replicating Factories

    Speaker_0 argues that superintelligence requires physical action, not just computation. He envisions a future where AI directs a profusion of specialized robots (drones, legged robots) and automated factories that build other factories. He likens this to von Neumann probes, suggesting the physical world will be increasingly directed by AI to support infrastructure like data centers.

  14. 56:17 – 59:57

    Self-replicating machines and human control

    Speaker_0 argues that self-replicating machines are a superior form for spreading through space compared to humanity. He emphasizes the need to ensure this development benefits humanity and keeps humans in control, rather than sidelining them.

Assets discussed

AssetKindStanceContext
OpenAIprivate companyneutralEven two years ago in 2024, when I was chief research officer at OpenAI, like the actual limiting factor was not the number of researchers we could hire.
Anthropicprivate companyneutralWell, you know, I think it could be either way, right? But then do anthropic, do OpenAI, do they have a comparative advantage in building out wet labs?
AppleAAPLequityneutralSo the iPhone, you know, famously designed in California, manufactured in China, it says that on the back of your iPhone.

“Stance” is how the speakers talked about the asset, not a recommendation.

Claims worth verifying

Bob McGrew was the Chief Research Officer at OpenAI.

Unidentified speaker (speaker_1) ·0:28

How to verify: Check Bob McGrew's LinkedIn profile or OpenAI's official press releases/archives for his title and tenure dates. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Bob McGrew was an early key executive at Palantir.

Unidentified speaker (speaker_1) ·0:28

How to verify: Check Palantir's SEC filings (e.g., S-1) for early employee lists or McGrew's LinkedIn profile. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Primary source

Automating researchers is a one-time boost, not a runaway self-improvement loop.

Unidentified speaker (UNKNOWN) ·3:16

How to verify: This is a theoretical claim; verify by analyzing the speaker's logic regarding compute and physical bottlenecks versus labor constraints in R&D. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

At OpenAI, the limiting factor for research was the amount of GPUs, not the number of researchers.

Unidentified speaker (speaker_0) ·4:16

How to verify: Verify by checking OpenAI's public statements on compute constraints or industry reports on AI training costs during 2024. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

iPhone 70 has 75% of its parts changed annually.

Unidentified speaker (speaker_0) ·14:37

How to verify: Verify by checking Apple's supply chain reports or industry analyses on iPhone component turnover rates. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Arda partnered with Anduril to build a software and AI platform for manufacturing engineers.

Unidentified speaker (UNKNOWN) ·25:53

How to verify: Check Arda's or Anduril's press releases, website, or news articles from late 2026/2026 confirming the partnership. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Arda reduced virtual build time for an actuator from days to 90 minutes.

Unidentified speaker (speaker_0) ·27:12

How to verify: Verify by checking Arda's case studies, press releases, or technical whitepapers detailing the performance metrics of their platform with Anduril. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Arda is working with Anduril on the Fury collaborative combat aircraft.

Unidentified speaker (UNKNOWN) ·27:44

How to verify: Verify by checking Anduril's product announcements or defense industry news regarding the Fury program and its supply chain partners. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

There are about a million people assembling iPhones in Asia.

Unidentified speaker (speaker_0) ·15:55

How to verify: Verify by checking labor statistics from Apple's supplier responsibility reports or industry analyses on iPhone manufacturing workforce. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

There are about 10,000 manufacturing engineers involved in iPhone production.

Unidentified speaker (speaker_0) ·16:15

How to verify: Verify by checking industry reports on manufacturing engineering workforce sizes in the consumer electronics sector. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Palantir developed the forward-deployed engineering model around 2005 for the intelligence community.

Unidentified speaker (speaker_0) ·45:35

How to verify: Palantir's public history and SEC filings (if public) or company press releases from 2026-2026 detailing their methodology and early government contracts. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Primary source

Bob McGrew worked with Ilya Sutskever at OpenAI and disagreed with his view that scaling training data was superior to learning from individual customers.

Unidentified speaker (speaker_0) ·30:57

How to verify: Bob McGrew's LinkedIn profile, OpenAI press releases from his tenure (2026-2026), and public interviews or articles featuring both McGrew and Sutskever discussing company strategy. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

OpenAI's robotics team spent two years teaching a Shadow Hand to solve a Rubik's Cube.

Unidentified speaker (speaker_0) ·36:39

How to verify: OpenAI's research blog posts or technical reports from the mid-to-late 2010s regarding their robotics projects, specifically those involving the Shadow Hand and dexterous manipulation. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Wojciech Zaremba was the head of the robotics team at OpenAI.

Unidentified speaker (speaker_0) ·36:19

How to verify: Wojciech Zaremba's LinkedIn profile, OpenAI organizational announcements, or academic publications listing his affiliation with OpenAI's robotics division during McGrew's tenure. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Palantir became a $300 billion company.

Unidentified speaker (UNKNOWN) ·50:46

How to verify: Palantir's stock price history and market capitalization data from financial news sources (e.g., Bloomberg, Yahoo Finance) between 2026 and 2026. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Bob McGrew taught himself deep learning by watching Andrej Karpathy's videos.

Unidentified speaker (speaker_0) ·35:43

How to verify: Andrej Karpathy's YouTube channel and blog posts from the period when McGrew was at OpenAI, and McGrew's own public statements or interviews confirming this self-education method. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Figure's model is called 'Fable'.

Unidentified speaker (UNKNOWN) ·41:48

How to verify: Figure AI's official website, press releases, or technical documentation identifying their base model or platform as 'Fable'. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Bob McGrew was a PhD student at Stanford's AI Lab from 2002 to 2005.

Unidentified speaker (UNKNOWN) ·34:23

How to verify: Bob McGrew's LinkedIn profile, Stanford University alumni records, or academic publications from that period listing his affiliation with Stanford AI Lab. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

AlexNet occurred around 2015.

Unidentified speaker (UNKNOWN) ·35:24

How to verify: Historical records of the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) results, where AlexNet won in 2026, and subsequent industry adoption timelines discussed in tech history sources. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Von Neumann proposed the concept of self-replicating probes.

Unidentified speaker (UNKNOWN) ·54:08

How to verify: Scientific literature on von Neumann probes, including John von Neumann's original theoretical works on self-replicating automata or subsequent papers citing his work in the context of space exploration. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Self-replicating machines are a better form factor for spreading between the stars than humanity.

Unidentified speaker (UNKNOWN) ·56:17

How to verify: Theoretical physics and aerospace engineering literature on von Neumann probes and interstellar colonization strategies; no single primary source, but checkable against established astrobiology and space exploration theories. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Self-replicating factories and AI suggest we should have encountered other intelligent life sooner than we have.

Unidentified speaker (speaker_2) ·57:12

How to verify: Review academic papers on the Fermi Paradox and the 'Great Filter' hypothesis, specifically those linking technological acceleration (AI/robotics) to faster interstellar expansion rates. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Civilizations spreading via self-replicating probes do so at the speed of light, meaning we only see them minutes or years before arrival.

Unidentified speaker (UNKNOWN) ·57:42

How to verify: Physics principles regarding the speed of light limit; verify against astrophysical models of signal propagation and probe travel times. This is a logical deduction from the speed of light constraint, not a new empirical fact. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Uncertainty in the Drake equation factors makes it plausible that humanity is the first civilization.

Unidentified speaker (UNKNOWN) ·58:47

How to verify: Review statistical analyses of the Drake equation, such as those by scholars like Robin Hanson or recent astrobiology studies (e.g., from the NASA Astrobiology Institute) that model probability distributions for the emergence of intelligence. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

We are unlikely to encounter alien civilizations; if we do, it will be shortly before their arrival.

Unidentified speaker (speaker_2) ·59:17

How to verify: This is a speculative claim/hypothesis. It can be checked against existing literature on the 'Dark Forest' theory (popularized by Cixin Liu) or other SETI risk assessment frameworks, but it is not an established fact. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify Anduril's public announcements and press releases regarding their partnership with Arda and the specific metrics of reduced build times.
  • Review Palantir's SEC filings and investor presentations to understand their forward-deployed engineering revenue model and customer concentration.
  • Check OpenAI's technical reports and Ilya Sutskever's public statements on data scaling strategies versus customer-specific fine-tuning.
  • Analyze robotics companies like Figure AI and General Robotics for their technical whitepapers on model generalization versus specialization.
  • Research the historical performance of neural networks versus SVMs in 2005 to verify the claim about the shift in AI confidence.
  • Investigate the Drake equation parameters and recent astrophysical data supporting the 'first civilization' hypothesis.
  • Look for primary sources on von Neumann probe feasibility and current self-replicating factory research in advanced manufacturing.
  • Check Arda's patent filings or technical documentation for their 'design for manufacturing' feedback loop implementation details.
Full transcriptexpand

0:00Unidentified speaker (speaker_0): If you look around outside, how different is anything from 10 years ago before AI happened? Not very different at all. The first time you were in your Waymo, you thought to yourself, this is amazing! Probably the third time you're in your Waymo, you're like, man, this traffic sucks. We need to break out of this pure focus on the digital world and go to the physical world. I don't think you are actually going to be able to get to superintelligence without changes in the physical world.

0:27Unidentified speaker (UNKNOWN): Hi

0:28Unidentified speaker (speaker_1): listeners, welcome back to No Priors. Today Aladdin I are here with Bob McGrew, the CEO and co-founder of Arta. Bob was previously an early Keeksek at Palantir and the chief research officer at OpenAI. We talk about making American manufacturing competitive with intelligence, why physical intelligence is necessary for superintelligence, and why the future is going to be machines that can make machines and explore the stars. Welcome, Bob.

0:55Unidentified speaker (UNKNOWN): Bob,

0:55Unidentified speaker (speaker_2): thanks so much for joining us, Sidana Priors.

0:57Unidentified speaker (speaker_0): Great to be here.

0:58Unidentified speaker (speaker_2): So you've had a ridiculous career. You were the chief research officer at OpenAI. So you drove all the sort of research directions for the company, or at least a lot of the agenda. You were at Palantir quite early and one of the main people there driving product and engineering. So you've had this amazing career spanning some of the most seminal moments of the digital world.

1:14Unidentified speaker (UNKNOWN): And

1:14Unidentified speaker (speaker_2): yet now you're working in the physical world. And it's right in the era when people are talking about RSI, which is recursive self-improvement, about how we will no longer have researchers and be displaced by code and AI and you know, AI building itself. Why do anything in the physical world right now as this digital revolution is happening?

1:31Unidentified speaker (speaker_0): Well, you know, fundamentally, I don't think you are actually going to be able to get to superintelligence without changes in the physical world. So, you know. the we've seen a huge amount of improvement from Chat GPT, from agents, you know, all of this is happening in our digital lives. And very little of that has actually affected the atoms of our lives. Like if you look around outside, how different is anything from 10 years ago before AI happened? You know, not very different at all. You know, even when you're doing something that is genuinely very different, maybe you're in a self-driving car, right? You're, you know, in a self-driving car, in your Waymo, going around San Francisco. And the first time you were in your Waymo, you thought to yourself, this is amazing. There's there's no one driving it. The like car is driving itself. And like probably the third time you're in your Waymo, you're like, traffic, man, this traffic sucks. Yeah. Right. Because just, you know, you have something, even you know, a Waymo is like

2:25Unidentified speaker (UNKNOWN): halfway towards the

2:27Unidentified speaker (speaker_0): physical world, but like you didn't change the car. Like the car doesn't fly. You're still bounded by all of the physical barriers that, you know, made the car trips around San Francisco pretty, you know, slow, even in the 1990s. Right. And so, you know, fundamentally, we need to break out of, you know, this pure focus on the digital world and go to the physical world. So

2:50Unidentified speaker (speaker_2): Nissan, you can't have super intelligence without changing the physical world. What does that actually mean? Yeah.

2:54Unidentified speaker (speaker_0): What does

2:54Unidentified speaker (speaker_2): that substantiate as?

2:55Unidentified speaker (speaker_0): So so maybe a good place to start here is I think there's this common perception andor there's a common argument. And I I've been hearing it since like 2018, which says that if you can just automate the work of the researcher, that you're going to be able to unleash a runaway acceleration, runaway self-improvement loop, and we're going gonna to have a singularity.

3:15Unidentified speaker (speaker_1): This is a hard takeoff.

3:16Unidentified speaker (UNKNOWN): Yeah,

3:16Unidentified speaker (speaker_0): hard takeoff. Exactly. And I think this is wrong. I think this argument is just straightforwardly wrong. So, you know, just to recapitulate the argument, it says, you know, okay, so we can automate 10% of a research time. That is gonna give, you know, that's gonna give some improvement. We automate 50% of a researcher's time. Now you have a 2x improvement. So now you're going 2x faster. And then you automate 90% of a researcher. researcher's time, you're going 10x faster. 99% automated, you're going 100x faster. And so by automating more and more of this job, you know, and eliminating this labor of the researcher, you're able to just, you know, zoom up and have runaway acceleration. But this is, this is this is straightforwardly wrong because researchers are not the bottleneck, right? Like the amount of time it takes in order for you to discover something is not primarily bounded by the number of researchers you have. Fundamentally, those researchers are running experiments. And so, you know, if you if you actually think about the the, you know, even today, right? Even two years ago in 2024, when I was chief research officer at OpenAI, like the actual limiting factor was not the number of researchers we could hire. It was the amount of GPUs that we had on hand. It was the physical amount of compute that we were able to put out there. And, you know, was the compute accumulated in the right places? Was it connected together with the right networks? And

4:36so you can, you know, completely automate researchers. This is a big effect, right? I'm not saying this is not a big deal. You know, maybe this is even maybe the ability to do automated research is going to be as impactful as the ability to go. from GPD 4.0 without reasoning to O3 with reasoning and tools. That was a huge one-time acceleration, but it's not RSI. So I think the right way to think about research is that, you know, fundamentally, there's a bunch of experiments. You have to run these experiments. And right now there's a lot of waste. You have to, you know, researchers have to do a lot of engineering. They have to pick the right experiments. They have to just, you know, go physically schedule them on the GPUs. And all of these lead to the GPUs being less effective at delivering new discoveries. And so, you know, the correct mental model is to look at the GPUs and to say, okay, well, you know, right now the GPUs are 35% of the way to

5:37optimal in order to be able to. deliver discoveries. And if we can automate more and more, maybe we can get them to 90% of the way to optimal.

5:45Unidentified speaker (UNKNOWN): But

5:45Unidentified speaker (speaker_0): the difference between going from 90 to 99% of optimal, it's actually not a big difference. And so we can automate researchers. It's a big one time benefit, but it doesn't kick in a self improvement cycle. Instead, it's one more, you know, S curve, one more big one one-time time improvement similar to research. reasoning that allows us to

6:06Unidentified speaker (UNKNOWN): uh

6:06Unidentified speaker (speaker_0): you know get closer to optimum. And then we have to go do something else. Yeah,

6:11Unidentified speaker (speaker_2): I feel like a lot of um AI driven research is basically stacking S curves of different types or AI development, I should say,

6:16Unidentified speaker (UNKNOWN): in

6:17Unidentified speaker (speaker_2): general. Uh that also translates into other domains, though. There's robotics, there's biology, there's a few other places where the physical world is really

6:24Unidentified speaker (UNKNOWN): uh,

6:25Unidentified speaker (speaker_2): you know, the primary limiter on. progress, right? You have to run the biology experiment, you have to build the,

6:30Unidentified speaker (UNKNOWN): you know,

6:31Unidentified speaker (speaker_2): the motors for the the sort of mechanical hand or whatever you're building. How do you think about the progress that the labs will drive versus startups will drive in the context of these other big markets where there's a strong research overlay or angle,

6:43Unidentified speaker (UNKNOWN): but

6:43Unidentified speaker (speaker_2): then there's sort of this physical world limitation that you have to overcome or at least optimize against in parallel? Like will it all will robotics and biology aggregate to the anthropics and opening eyes and Googles of the world and their subsidiaries like isomorphic, or do you think there'll be like a bunch of startups that drive this sort of progress?

7:00Unidentified speaker (speaker_0): Yeah. Before we talk about market structure, I think like let's let's think about what the actual maybe the economic structure of this is. You know, the labs, if if we think about what the labs are doing, they're really producing cheap intelligence. So what are the complements to intelligence? If we think about biology, or again, we think about running these experiments for superintelligence research, you know, the intelligence in terms of choosing which experiments to run, generating hypotheses, it's really a complement to the physical layer of having the wet labs, of you know, laying out the GPUs. And so, you know, when hypotheses become cheap, then it's actually the physical part that becomes scarce. And so, you know, you know, if if you're if you get better and better at generating hypotheses, you're just gonna bang harder and harder against the limited number of wet labs that you have. And so then, okay, so go back to the market structure. Well, you know, I think it could be either way, right? But then do anthropic, do OpenAI, do they have a comparative advantage in building out wet labs? I think they probably don't.

8:07Unidentified speaker (speaker_1): Did you consider doing other working in other domains before you decide to go work on manufacturing and robotics?

8:14Unidentified speaker (speaker_0): I did. Yeah, yeah. So when I left OpenAI, I think the one thing I knew was I didn't want to start another frontier lab. Like, in fact, maybe I wanted to be the only person that's ever left OpenAI and not started a new frontier lab. And you know, I'd done that and I wanted to do something new. And I was basically, I was, I was captivated by this question that the models are so powerful, right? And they're really powerful now, but they were really powerful then too. And yet they don't seem to be making that big a difference in the world. And

8:44Unidentified speaker (UNKNOWN): you know,

8:45Unidentified speaker (speaker_1): people people love to talk about abundance as a concept in Silicon Valley. And I think it's a very interesting one because it's it's a very physical word, right? When you think about abundance, you're like, I have an abundance of food or housing or family or something like that. Like the way we experience the world around us is very physical. And I don't think anybody is like, I necessarily have an abundance of digital work.

9:08Unidentified speaker (speaker_0): Yeah. And and in and in and order to actually get something done, you know, in economics, I think it's called like an O-ring production function. but like you have to do all the things. And so if we're able to speed up the purely digital process of building housing,

9:21Unidentified speaker (UNKNOWN): um,

9:21Unidentified speaker (speaker_0): well, that's that turns out not to be very important. Right. Actually, the hard part of building housing is getting the permits, which no amount of agents are gonna help you do, right? But also you have to, you know, do the construction work.

9:32Unidentified speaker (UNKNOWN): Um,

9:33Unidentified speaker (speaker_0): you know, I have a I have a mental model that I think about for what it looks like to

9:38Unidentified speaker (UNKNOWN): you

9:39Unidentified speaker (speaker_0): know the the impact of superintelligence, where people say, okay, well, if we can just have sufficiently smart agents, then you know, we're gonna have Dyson spheres in just a few years. And so, like, here's a here's a thought experiment. Is, you know, actually, let's say we're blocked on building data centers. Like we want to build a lot of data centers. Like, we need to go build these data centers. We're slow. What is it? Well, we have a shortage of bulldozers. No problem, right? Well, the A. AI knows what needs to be done. It's going to go like build a bulldozer factory. So now we got to build a lot more bulldozers. Well, great. What's the first step in building a bulldozer factory? Well, you got to clear the ground. What do you need? Well, you need more bulldozers.

10:15Unidentified speaker (speaker_2): Well, I think, yeah, that's interesting. I guess the the comment I was going to make really quick on all this is I totally get your point on construction. I wonder for certain domains like biology, chemistry, materials, et cetera, if you're able to build a sufficiently strong internal model that can actually simulate a lot of these things well enough. Protein folding is a great example where you really don't have to go and solve all these crystal structures anymore. You don't actually have to do the experiment anymore because the in silico prediction is so good. And so I wonder if you actually fragment what you can do in the physical world into two camps. There's a camp where you have to literally go and build the bulldozer and figure out, you know, or build the manufacturing capability to build the next manufacturing capability or whatever it is. And then there's the ones where you can actually simulate the experiment to the point where you don't need the wet lab anymore. Or you minimize the number of things you need to do in the wet lab, or you need to do in the materials shelf, or you need to do in the sort of physics lab. And so I just wonder if it's a little bit if if it's if all these things aren't exactly the same. There's two different types of problems that we're dealing with in within the physical world.

11:10Unidentified speaker (speaker_0): Yeah, that's really interesting. I think they're going to feel the same.

11:14Unidentified speaker (speaker_2): And the

11:14Unidentified speaker (speaker_0): reason. they're going to feel the same is because of the tremendous speed with which we can we're going to be able to do anything digital. So let's go to wet labs, for example. Let's say that we build all amazing, you know, digital simulations of proteins and of how they interact in a cell. And we're able to do this. And suddenly we are able to predict 10 times better how good what's going to happen with our experiments. And so now, in order to get to a discovery, we need to run one tenth as many experiments as we needed to run. Well, we're going to be that period of time in which we are blocked on the digital world instead of the physical world. We're very fast and moving in the digital world. So we're going to eat up that acceleration. It's similar to all the other things. It's going to be a one one-time time acceleration, but then we're going to be blocked on the physical. world again. And then now we're 10 times better, but we still have to go run those experiments. I think there's an analogy that I remember from OpenAI where, you know, I was

12:14the chief research officer, Ilya was the chief scientist. And, you know, especially early on, I would I would read these papers and I'd be like, oh, these methods, they're so cool. And I'd go to Ilya and I'd be

12:23Unidentified speaker (UNKNOWN): like,

12:23Unidentified speaker (speaker_0): oh, they have this idea and it works this way. And Ilya would like list. listen patiently for me to explain the paper to him. And at the end, he would say, So what are the results? Because, you know, even Ilya, right? It's like, yeah, sure, this it's a cool idea. It could work, it could not work. But at the end of the day, you have to run the experiments. You know, I think if deep learning, which is has no physical, you know, it's like a purely mental, it's a purely in some sense a mathematical. mathematical science. Actually, it's an experimental science. Right. And even Ilya, you know, he can tell you, yeah, this is a bad idea, don't even try it. But there's this huge number of ideas that are that are actually very good. And just applying a lot more intelligence to figure out like, is it a good idea or is it not a good idea isn't gonna help you. You have to go run the experiment. And then maybe you do were able to find proxies for this, which is great. As we find those proxies, then you know, then we we scale up to accomplish that. And then again, we're we're

13:21Unidentified speaker (UNKNOWN): banging hard on the physical world.

13:22Unidentified speaker (speaker_2): Tell us more about what Arta is doing then specifically. Like how is Arta approaching the physical world? What are some of the products or areas that you're focused on?

13:29Unidentified speaker (speaker_0): Yeah. So so let me start by by talking about the problem we're trying to solve. So after I left OpenAI,

13:35Unidentified speaker (UNKNOWN): um,

13:35Unidentified speaker (speaker_0): you know, I I sort of noticed that we are very good in America at designing products, right? So the iPhone, you know, famously designed in California, manufactured in China, it says that on the back of your iPhone.

13:49Unidentified speaker (UNKNOWN): Um,

13:49Unidentified speaker (speaker_0): you know, we're amazing at designing defense materials, uh, but we don't manufacture very much of it here. And so I went around, spent a year talking to manufacturing leaders and asking them, like, what's going on here? And, you know, fundamentally, it comes down to fundamentally it comes down to labor costs. And so if you want to build something in America, one possibility is that you're in defense, in which case you can charge whatever you want. And if you're if it works, the government will buy it. So that's great. But if you're not in defense, then you need to build an automated line if you want to be able to build something cost competitively in the US. We're building all we're designing all these innovative products One part of designing an innovative product is that the product itself changes. And so famously, you know, each year the iPhone 70 has 75% of the parts in it change. Only about 25% stay the same. So that 25% is a highly automated line. The other 75% is assembled by hand. And the reason for that is that if you were to try to build automation that can go assemble this, that automation is very specific. The way we build automation today, it's like

14:57a bespoke machine that solves one particular physical problem. And that machine doesn't generalize from one phone to another.

15:07Unidentified speaker (UNKNOWN): And

15:07Unidentified speaker (speaker_0): it doesn't even generalize from one generation of the iPhone to another. So going back to what we can build in America. well, if you're building a car, right, great. Build a highly automated facility, spend you know, a couple of years and $200 million to make it happen, and then produce that same car with minor cosmetic updates for seven years. The economics work great. But if you want to do that for an iPhone, first off, you have to have you have to go from about 25% automated to about 90% automated for it to pencil. And then how are you going to be able to replace all of that automation? Like that's just a huge, incredible cost that you're not going to be able to amortize unless the automation becomes more flexible. And then there's a second piece of it, which is that who is actually going to be designing this automation? There are about a million people assembling iPhones or components in Asia. There are alongside that, about 10,000 manufacturing engineers who are figuring out what is the manufacturing process that we need to do. How do we stand up these assembly lines? How do we spin down these assembly lines when it comes time to spin up the next product?

16:15And so that's fundamentally the problem we're trying to solve is how can you make it It's as easy to turn over an automated line as it is to change? and retrain workers in a human plant. So you can retrain a human worker in a day. Practically, manufacturers think they're extremely agile if they can bring a design change into production even on a manual line in seven days. We would like to be able to build a system where we are building a system that allows you to upload your product design into our cloud and then design and ultimately reconfigure an automated line in seven days.

16:56Unidentified speaker (speaker_1): Putting aside the fact that everybody would like to be able to make whatever they want on demand as quickly as possible with that sort of pace of innovation, like what are the types of products and customers where this matters the most from a business perspective, where that high mix and speed to bringing up the line is really important to them. Because you kind of said, like for car, it's kind of okay to have the same car for seven years.

17:18Unidentified speaker (speaker_0): It basically is anywhere that you are doing highly innovative, frequently changing manufacturing, or where you have a lot of variants. Defense is, I think, a very interesting example. It's very close to my heart as you know a Palantir veteran. And, you know, in Ukraine, for example, you are now going, you're now spending only a few weeks from change. changes in the battlefield into physical changes in the design that then need to be pushed out back onto the battlefield. That is a very short amount of time. And it's driven by these competitive pressures, right, between Ukraine and Russia. Well because

17:53Unidentified speaker (speaker_1): if if your adversary can turn that around faster than you can.

17:56Unidentified speaker (speaker_0): Exactly. Well, you know, there are also competitive pressures, although less violent ones, between Google and Apple. or, you know, the Android ecosystem and the i iOS ecosystem. And so, you know, as you are uh building these larger and larger markets and these more and more innovative products, what you want to be able to do is just crank up the pace of innovation. And so consumer electronics is a great example here. A lot of things that go into data centers need to be changed very frequently because. chips change very frequently. I think that's those are the kinds of industries: defense, consumer electronics, the data centers and data center supply chain, where you see a lot of need to be able to turn things over quickly and where as a result, almost everything is built by hand. And

18:42Unidentified speaker (speaker_2): when you talk about automating these lines, is that humanoid robotics? Is it just existing manufacturing technology repositioned? How do you think about the form that's going to take?

18:51Unidentified speaker (speaker_0): Yeah, I mean, the the the true answer is that it's all of the above. But practically, what does it look like? Um,

18:58Unidentified speaker (UNKNOWN): you know,

18:59Unidentified speaker (speaker_0): I and I I would I think there's the the most obvious thing is that you want to be able to have reconfigurable robotic work cells that you can immediately flow a design change into an actual change that's happening uh in the factory. That's a hard problem. Obviously, there's a lot of people working on different variants of that.

19:19Unidentified speaker (speaker_2): The reason I ask about it is I feel like there's two theories of the world, and I'm not saying they're incompatible. One theory of the world is everything is derived designed around people right now in terms of manufacturing lines, et cetera. And so therefore the most natural thing to do is just copy a person, which means have a humanoid robot. The other viewpoint is, well, there's lots of these um sort of very common manufacturing processes that you need to do. A lot of them are done by hand because to your point, it just it's a faster turn if you need to turn something around quickly. And those things could just be made into sort of stepwise sequential things like what we've done in automotive for welding arms or other things like that.

19:51Unidentified speaker (UNKNOWN): And

19:51Unidentified speaker (speaker_2): so I'm a little bit curious like where on that spectrum you think we are in the near term versus the long term.

19:55Unidentified speaker (UNKNOWN): Yeah.

19:55Unidentified speaker (speaker_0): So I so I would I would come in with two ideas. So first off, let's talk about the form, form factor. So for the form factor, given what I just told you. we are trying to replace humans, right? We're not trying to replace additional forms of automated lines. We don't need to work in automotive where they have these like extremely large Baroque robotics. Um, I think you know, two arms and a camera is enough. In a lot of cases, actually, one arm and a camera is enough. And maybe you need some more, some better sensing. Um, you know, uh so from a form factor perspective, these lines, you know, you have a human standing there all the time. You don't need a humanoid robot. And

20:30Unidentified speaker (speaker_2): does the capability

20:30Unidentified speaker (UNKNOWN): of the arms

20:31Unidentified speaker (speaker_2): have to change dramatically? Because people always talk about the sensitivity and fidelity of existing robotic arms versus humans and all the rest, like how much progress actually has to be made in terms of the component tree for the arm versus hey, it's actually ready to go and it's just all about the software.

20:44Unidentified speaker (speaker_0): Automated existing automated lines are able to build all of all of the products that we're talking about, or at least like 90% of them. You know, yes, wire harnessing is hard. But you know, if you go look at something like the iPhone and you ask what is the barrier to doing the automation, it is not that robot arms, you know, can't you know sufficiently screw in every screw and you know fiddle with every everything. That kind of sensitivity already exists. It's the economics because those arms and you know that those they call them capex, capital expenditures, those. machines have to be redesigned and rebuilt every year. So I think they

21:19Unidentified speaker (UNKNOWN): can do

21:20Unidentified speaker (speaker_1): it, it's just too expensive.

21:21Unidentified speaker (UNKNOWN): They can do it, it's just too expensive.

21:23Unidentified speaker (speaker_0): Yeah. And so I think there you want to actually switch from thinking about the embodiment to thinking about the intelligence layer. So for the intelligence layer, I would distinguish robotics as you know, some robotics problems are open world and some robotics problems are open. closed world. So if you were trying to clean a bedroom, that is an open world robotics problem. You have never seen the items here, you know, in a lot of ways that we think about this. When you're deployed there, you've never been in this bedroom before. You're never gonna go back into this bedroom. Um, it's this like very long horizon planning problem. Um, that's very hard. And it's very suited to, you know, a pure you want to think about this as as you know, the the intelligence you need is very similar to human intelligence. A closed world

22:07Unidentified speaker (UNKNOWN): um

22:07Unidentified speaker (speaker_0): is one in which you actually control everything and you're doing something very repetitively and there's no novelty that's being introduced. Assembly lines are a closed world. That makes sense because we see that we're actually able to use sort of traditional automation approaches in order to automate them. And I think it's worth talking a little bit about what those look like. So, you know, suppose that I am trying to

22:30Unidentified speaker (UNKNOWN): uh

22:31Unidentified speaker (speaker_0): solve some particular problem with with automation.

22:35Unidentified speaker (UNKNOWN): Um,

22:35Unidentified speaker (speaker_0): again, traditional robotic automation. What they end up doing, I think the best analogy for this is a Rube Goldberg machine. So I I had to build one of these with my son who's who's 13 for school, you know, and you're you're like, okay, well, you know, the ball rolls down the hill. and it hits the pencil and the pencil like pops the balloon. And, you know, instead of that, it's actually you have a chute and you know, the the nut goes down the chute, and you know, there's a cup and the cup is vibrating, and because it's vibrating, uh, it the nut drops out and it lands on a fixture.

23:07Unidentified speaker (speaker_2): Yeah.

23:08Unidentified speaker (speaker_0): And the fixture locates it precisely. So now there's no, you know, uncertainty about where it is. And a robot arm without any perception whatsoever can come and pick up the nut. And then they can go take it to the bolt that you're screwing the nut on, which is also on a fixture, is also precisely located. And then it can screw that on.

23:29Unidentified speaker (UNKNOWN): And

23:29Unidentified speaker (speaker_0): it's

23:29Unidentified speaker (speaker_1): basically deterministic. It's moving

23:31Unidentified speaker (UNKNOWN): the coordinate

23:32Unidentified speaker (speaker_1): to this coordinate.

23:33Unidentified speaker (speaker_0): Yeah. So what's hard about this, right? Like why, what, why is why is this interesting? Like, why does this take six months in order to turn around? Well, there's two things you have to do. One is you have made a machine.

23:43Unidentified speaker (speaker_2): Yeah.

23:43Unidentified speaker (speaker_0): Right. So someone needs to go design that machine. That is not an easy process. That is a mechanical engineering task. It is a manufacturing engineering task.

23:53Unidentified speaker (UNKNOWN): Um,

23:54Unidentified speaker (speaker_0): well, we now have AI that can design machines. So that's one way we can make it faster. The other thing, uh, which you know really hit. home to me as I was doing this Rube Goldberg project is like,

24:06Unidentified speaker (UNKNOWN): you know,

24:06Unidentified speaker (speaker_0): half the work is in the design, the other half is in the calibration. Right. So, you know, you got to make sure the ball hits the pencil exactly right so it pops the balloon.

24:15Unidentified speaker (UNKNOWN): Um,

24:16Unidentified speaker (speaker_0): you know, similarly, you know, if the nut is, you know, a tenth of a millimeter off, it may not end up in the right place on the fixture. And then, you know, the arm may not be able to uh actually, you know, place it in the correct place on the device.

24:34Unidentified speaker (UNKNOWN): That

24:34Unidentified speaker (speaker_1): sounds like a relatively simple RL problem. No?

24:37Unidentified speaker (speaker_0): Well, I think there's a lot of, you know, this is, I think, where something like agentic robotics becomes an interesting way to look at the world because, you know, we are no longer in this open world, right? We're now in this extremely closed world doing a set of very specific uh motions. And you can bring in perception. And now, you know, it's you know, if the nut's a little bit off, like you don't need to calibrate it quite so precisely, you can go pick it up and you can bring in RL and manipulation. And now, you know, maybe the nut's in the wrong place on the hand, for example. You know, the the device is always gonna going to be in a fixture, but but now you're you're using actual manipulation. in order to screw this in. And maybe you can start by automating the simpler kinds of stations. And then over time you get into the ones that are really wacky where you've got, you know, elastic things that need to be stretched over a gasket. And again, though, speaking of closed world, in an open world, you know, maybe you know you miss a couple of things in the bedroom and that's fine. In a closed world, actually, like 99 or

25:3999.9% of the time, you got to go construct the device that you are constructing.

25:44Unidentified speaker (speaker_1): How tall is the the hill to climb to creating a great manufacturing engineer? Like how deep is that field? How far do you think you are?

25:53Unidentified speaker (UNKNOWN): You

25:53Unidentified speaker (speaker_0): know, the starting point for us, you know, we've been talking about designing automated lines. We've spent the last year partnered with Andorral Androil on building a software. and AI platform that allows manufacturing engineers to design lines. Now, you know, Android's a defense manufacturer. You know, they don't need to have automated lines, and a lot of the products they have aren't high enough rate yet to support traditional automated lines. And so what we've done is, you know, basically, so we went in there a year ago. We actually talked to them the day before we incorporated the company and they told us about their challenges. They talked about the need for a single source of truth for manufacturing engineering, where one place where you actually encapsulated the manufacturing process and the manufacturing plan and a reconfigurable factory that would allow you to change the manufacturing process and then also spin up and spin down different lines based on demand. And so we adopted those as sort of our North stars for automating manufacturing engineering. And, you know, we got access to their systems in January. In April, we had our first user. So he was able to take an actuator that he was trying to build. And normally that would have taken him a few days in order to do the virtual build for this. And he was able

27:12to do that in 90 minutes using our software. And that was, you know, the very first thing we did. And since, you know, and that that worked out really, that saved them a lot of time and they were very excited. And so since then, we have been moving across all of their product, we're now deployed across all of their product divisions. Uh, we've done work with them, you know, on everything from drones to missiles. Uh, you know, we're currently working with them on the Fury uh collaborative combat air. airplane. So we're taking this entire airplane and, you know,

27:44Unidentified speaker (UNKNOWN): uh

27:44Unidentified speaker (speaker_0): taking that manufacturing process and making that legible to AI. And then,

27:51Unidentified speaker (UNKNOWN): you

27:51Unidentified speaker (speaker_0): know, a lot of what we're doing is basically you encapsulate that manufacturing process. Now it's not scattered in a hundred 100 different documents and in a hundred different places. And so you can have an AI agent that actually is able to access. add value. So, you know, going back to what does it take to be to your question, to what does it take in order to truly, you know, build an automated manufacturing engineer? We are lucky that we can start by saying, okay, we want to be cursor here effectively, noting that cursor basically said, okay, we're going to take this existing IDE and we're going gonna to add an agent. There's no IDE for manufacturing engineers. So first we had to build that so that there's a true data structure, there's equivalent of of effectively Git that you can work with. And then we can apply the agent on top of that. And then now that we own and and well now that we control uh in a in a specific way the manufacturing process, we can, you know, move left and right of that. So now you can say, okay, well, this design you know, is manufactured this way. If we wanted to manufacture it more cheaply, how would we have to change the design? So we can send design for manufacturing feedback back into the

28:59design engineering, which is again a role of the manufacturing engineers, but they're able to use our agent to do that much faster and more effectively. And similarly, we can then say, okay, we can move right and say, if we want to have an automated station, how do we design an automated station that can produce one of these, you know, that can do one of these manufacturing steps?

29:23Unidentified speaker (speaker_2): Yeah, this is um definitely the most pragmatic approach to this world that I've seen. And I feel like there's been a lot of iteration over time. I remember like a decade ago funding things that were going to be like Git for hardware or other things that kind of impinge on this in terms of how do you store the knowledge and then how do you actually act on it? And it feels like the unlock is really happening with AI. One thing that I think is really pragmatic about what you've already what you've described to me in the past in terms of the approach that you're taking is you're also not going and trying to build massive models of your own.

29:50Unidentified speaker (UNKNOWN): And

29:50Unidentified speaker (speaker_2): instead, you're saying, how do we actually use things that are a bit off the shelf from an AI perspective to be able to really make this streamlined and fast and efficient and cost effective? Could you talk more about that choice of like building your own models versus

30:03Unidentified speaker (UNKNOWN): uh

30:03Unidentified speaker (speaker_2): trying to use a broader suite of things against this?

30:05Unidentified speaker (speaker_0): Yeah. Well, I would start, but just by the from the perspective of simplicity. Right. So, you know, when you want to go solve a problem, you want to use the simplest tool for the job. And, you know, maybe even to step back. So, you know, I had you know one career at Palantir, one career at OpenAI. At Palantir, we were relentlessly problem problem-focused. focused. You know, we wanted to build the enterprise platform for the intelligence community, which ultimately became the enterprise platform for all enterprises. And, you know, we didn't care what technology we needed. And we had to solve a lot of very hard technological problems, but we used the forward-deployed engineering approach. We talked to our customers, we heard what they needed, we brought that back into the product, and we just relentlessly expanded that. But by focusing on the problem, at OpenAI, we took the option. opposite approach. We had no idea what any of this would be used

30:55Unidentified speaker (UNKNOWN): for. Um, you

30:57Unidentified speaker (speaker_0): know, we I remember having a conversation with Ilya once where uh we had just built GPT 3 and you know, we we were like, this is pretty good, but it needs to be better. And Ilya said, Well, Bob, how do you actually build a company? Like,

31:09Unidentified speaker (UNKNOWN): what

31:09Unidentified speaker (speaker_0): would we need to do to make this product? I said, Well, you know, there's a standard approach that everybody in startups knows, which is you pick someone who, you know, has a real problem, you actually solve their problem, and then you expand from there. And Nilya said, Ah, yeah, that's very interesting. Yeah.

31:24Unidentified speaker (speaker_2): No,

31:24Unidentified speaker (speaker_0): that's not going to work for us.

31:25Unidentified speaker (UNKNOWN): And

31:26Unidentified speaker (speaker_0): I said, What? What well

31:27Unidentified speaker (UNKNOWN): I

31:27Unidentified speaker (speaker_0): mean, thanks for thanks for asking, but what why won't this work? And he's like, Well, look, we could learn something by talking to one person, but instead we have the ability to just, you know, 10x our training data, and then we'll learn so much more from everybody, and that's much more than what we can learn from one customer. So I think that's that's the justification for why you want to take a technology first approach to your market. Because fundamentally, this technology has its own way of finding out about the world. Well, at Arta, we're taking a technology first approach, we're taking a problem first approach. And so why are we not training models? I've trained models before. We don't need to do it. Um, for the uh the first step of understanding the manufacturing process, designing a line from that, that's something that we see that frontier models can do. Uh, when you start going into something like robotics, then um frontier models can do some of it.

32:21Unidentified speaker (UNKNOWN): Um,

32:22Unidentified speaker (speaker_0): but what what we see is that basically the frontier models may work best as an organization. orchestrator. And in fact, that's that's actually a lot of what we're doing with uh manufacturing engineering is that we have built um specialized algorithms, we use motion planning,

32:37Unidentified speaker (UNKNOWN): um

32:38Unidentified speaker (speaker_0): we have different tools that you can use to inspect the CAD. The frontier model is not adding all the value and they're an orchestrator of all of these different tools.

32:48Unidentified speaker (UNKNOWN): Yeah,

32:48Unidentified speaker (speaker_2): you actually see this a lot in the SaaS world too, where you use the SOTA model as the orchestrator, and then you actually have fine-tuned models or specific things that are very specialized for the application area. So it's interesting to see that it's also happened here.

32:58Unidentified speaker (speaker_0): Yeah. And and I I think it's just like a good general purpose mechanism for thinking about things. Ultimately, if you want to have the most intelligent model, there will be, you know, one a small number of, you know, very capital capital-intensive, intensive, very smart models. Like, you know, we I don't think you're gonna have specialist models unless you're thinking about cost. But when you're trying to, when you're not a frontier lab, it's actually extremely approachable to make you know small special purpose models that are better than you know what the off-the-shelf frontier models can do, at least for now. And training one of those models versus like writing a software package versus writing a tool, like those are all kind of the same. action.

33:39Unidentified speaker (speaker_2): How did you um end up at OpenAI coming from Palantir? Because to your point, you came from such a pragmatic world. And I remember visiting OpenAI in 2018, 2019. And at the time there was a handful of like robotic arms that weren't doing much in the office. And there was a bunch of um the a lot of the focus, I think, was on video games and sort of building beating video games initially. And then it and then you guys started on the transformer and sort of GPT train. But for a while it was kind of this. very broad based research approach. Let's try a couple things. At least that's how I saw it as an external viewer.

34:09Unidentified speaker (UNKNOWN): Was

34:09Unidentified speaker (speaker_2): that a big transition? Why even join an organization that was so radically different from what you were used to?

34:13Unidentified speaker (speaker_0): Yeah. So so I would I would actually start the answer to this uh 10 years prior. So I was doing my PhD at Stanford in the AI lab.

34:23Unidentified speaker (UNKNOWN): Um,

34:23Unidentified speaker (speaker_0): and you know, this from 2002 until I dropped. out in 2005 to join Palantir. And I remember vividly sitting around with my research group making fun of people who thought that neural networks would ever work.

34:37Unidentified speaker (speaker_2): Uh-huh.

34:37Unidentified speaker (speaker_0): Because, you know, there was no, you know, proof that they would work. Uh, you know, unlike SVMs or regression, which have clear provable Improvable guarantees.

34:46Unidentified speaker (UNKNOWN): Uh,

34:46Unidentified speaker (speaker_0): and even if they did work, you couldn't figure out why they worked. And so you know, like why, why, why would you ever use these things when obviously, obviously, all these other things work. And so I left Palantir because I left uh and went to Palantir because I thought that AI wasn't really happening back in 2005. So spent 10 years at Palantir, run engineering and product management there. And, you know, around 2015, I'm looking around. It's been a few years since Alex. Alexnet happened. Uh, vision models are beginning to really work. Palantir has grown tremendously, and I'm, you know, thinking maybe it's time for me to move on. And,

35:24Unidentified speaker (UNKNOWN): you

35:25Unidentified speaker (speaker_0): know, I thought to myself, like, maybe AI is really working now. And where is it going to be an application of all the vision models that people are were doing? thinking, Well, sure, you can go recognize people's faces and photos. And if you want that, you should work at Facebook, not start a company. But robotics seemed like maybe this would be the place.

35:42Unidentified speaker (UNKNOWN): Um,

35:43Unidentified speaker (speaker_0): I was very wrong about that. I spent a year and a half trying to get this to work. I taught myself, you know, deep learning from watching Andre Karpathi's videos. Uh, you know, got to the point of teaching a robot to play checkers from vision. And having done that, I realized this is way too hard. This is not the right time for a startup. And so, uh I knew some people and uh you know got introduced to this nonprofit full of smart people that had the you know somewhat dramatic goal of building AGI. And I thought, well, this isn't a I I I'm not gonna start a company. Why don't I go work for a nonprofit for a while? And so I started there. And, you know, I didn't think I would stay for years. I expected I would be there for a couple years. of I would learn from people like Ilya or Wojciech Zaremba or Andre Karpathi, who were there at the time. And I ended up joining the robotics team. And the goal for the robotics team was to show cool advances in robotics in a way that demonstrated that deep learning was actually a real thing that people should pay attention to, that it might in fact lead to AGI. And so we settled. on this project of have a humanoid robot hand, a shadow hand, uh be able to solve Rubik's Cube.

36:57Unidentified speaker (UNKNOWN): And

36:58Unidentified speaker (speaker_0): the reason for that, the way the story behind this is that Wojciech Saremba, who's one of the co founders of OpenAI, but also the head of the robotics team, went around at a robotics conference and asked people what would just be absolutely mind blowing to you if we solved

37:10Unidentified speaker (UNKNOWN): it.

37:10Unidentified speaker (speaker_2): Yeah.

37:11Unidentified speaker (speaker_0): And, you know, people said, well, either pick in place that can pick and place RB. arbitrary objects or or doing anything with a robot hand. And what would be the most impressive thing you could do with a robot hand?

37:21Unidentified speaker (UNKNOWN): I don't know,

37:21Unidentified speaker (speaker_0): solve Rubik's Cube.

37:22Unidentified speaker (speaker_2): Yeah.

37:23Unidentified speaker (speaker_0): And so we spent two years for that. And it actually worked. Um, although I I will, I'm now permanently not a believer in robotic hands.

37:32Unidentified speaker (speaker_2): Yeah.

37:32Unidentified speaker (speaker_0): Uh, after the experience of trying to make one do anything.

37:35Unidentified speaker (UNKNOWN): I hear

37:35Unidentified speaker (speaker_0): the I hear the engineering's better now, but I'd I probably ruin ruined for life. If

37:38Unidentified speaker (speaker_1): you fast forward. all the way to this point where Astra and Fable look like they're showing, you know, signs of life that excite or freak out the robotics community in terms of actually being able to generate policy to do things in the real world from a general model. Like, what do you make of that? Is it is that like the end of people training um models specifically in general robotics?

38:01Unidentified speaker (UNKNOWN): Well,

38:02Unidentified speaker (speaker_0): you know, the the end of people training models probably depends on venture capital. So, you know, that's more a question for you guys than for me. But like, why does this work, right? Like, yes, it's the bitter lesson. But fundamentally, you know, the challenge in robotics has always been that the amount of data that you have to work with is very small. And so you would see companies that, you know, we're going to solve robotics by teleoperation and we're gonna going build to giant teleoperation. labs. And, you know, then it's like, oh, well, we can use these UMIs, uh, you know, little devices like chopsticks that record, you know, people's operations. And then I can collect data 10 10x more cheaply. All of these are extremely scarce sources of data. I think about it like if we were trying to train GPD4, but we were doing it on top of, you know, if we had to pay us pay scale AI to go have humans write all of the data. like we would never have really been able to train GPD4. But fundamentally, you don't need to train the model on exactly what you're deploying on. What you need to do is generate really good representations of the world. And then you can fine-tune the models towards a particular problem. And so, you know, where do those representations of the world come from? They come from predicting the last few. significant bits of you know some sort of real world data that you're dealing with, which could be

39:21internet videos. Um, I'm sure there is, in fact, a tremendous amount of robot data that has been brought in, but I also suspect it's very small compared to the fraction of overall data in a model like Astra or Fable. And so as you know, the models progress, as the data gets better, as the algorithm algorithmic improvements change, um, the model becomes better at sort of eking out those last few bits of randomness and being able to predict those. And suddenly now it's able to predict events in the physical world. And once you have those really good representations, then you can go target them at robotics. And suddenly robotics begins to work. I suspect, you know, the there the for a long time there was this

40:02Unidentified speaker (UNKNOWN): theory

40:02Unidentified speaker (speaker_0): of. you know, emergent capabilities that, you know, I remember watching on Twitter sometime where someone said, Oh, you know, I watched GBD4 and it can go debug, it can like iterate with a software, you know, uh you you know, you can run this program, you can it can, you know, look at the output and it can go fix bugs, this crazy emergent capability. And I'm like, well, Jerry spent like two years working on that. Right. That's like not as emergent as you think it is. Yeah. And so, you know, yes, you know, we know that there are robotics teams at OpenAI, there are robotics teams at Anthropic that are doing this. But my guess is that the actual motor of progress here is not that they collected a lot more robot data than the next people, but that the models were able to finally, you know, get down to the last bits. of the physical data that they already had.

40:53Unidentified speaker (speaker_1): What do you think that means for the overall robotics ecosystem? Do you expect like maybe that's also a question for entrepreneurs and venture capitalists, but uh what do you expect the player landscape to look like? When do you expect people to have robots in different use cases?

41:07Unidentified speaker (speaker_0): Yeah, I think this really democratizes it. Um I I think there's an interesting tension here that I don't know how to resolve with which is, you know, for digital tasks, few shot learning works very well. And also, you know, you it's really easy to collect data that you can actually that that's similar in format to the data you're actually using. In robotics, you know, again, because of my suspicion that a lot of the value is not coming from the robot data in these models, um I've always believed that you're probably going to want to be able to fine-tune. fine tune

41:39Unidentified speaker (UNKNOWN): in

41:39Unidentified speaker (speaker_0): order to be able to work in any particular area. And so I think there's a tension here between something like Astra or Fable.

41:48Unidentified speaker (UNKNOWN): You know,

41:48Unidentified speaker (speaker_0): it's very good out of the box.

41:50Unidentified speaker (UNKNOWN): Um,

41:51Unidentified speaker (speaker_0): it's also slow. That can be overcome, but you can't fine tune it.

41:55Unidentified speaker (UNKNOWN): So

41:55Unidentified speaker (speaker_0): it's very hard to fine tune effectively on a specific use case. And so I think that's an opportunity for

42:01Unidentified speaker (UNKNOWN): uh

42:02Unidentified speaker (speaker_0): other. robotics players. And I think if you if you look at Pi or you look at generalists, like you can see them, you know, using this opportunity that they have to take data and use it to fine-tune against a specific use case. And so, you know, to you know, the bitter lesson is very strong. And also the idea of training on the test set is very strong. Yeah. Right. Both of those are very powerful forces. And I think we're gonna going to have to see where those come. in.

42:27Unidentified speaker (UNKNOWN): Um,

42:28Unidentified speaker (speaker_0): I do think this this does sort of leaping out of that frame though, is that agentic robotics, I think, is an interesting way, like sort of in between these two things, where um you can think of these models as just one more tool

42:42Unidentified speaker (speaker_2): that

42:42Unidentified speaker (speaker_0): the agents have. If you're doing something like we are, where you're working with the closed world, then you have a lot of different tools. Now you, you know. now if there's one thing that requires an incredible amount of desk dexterity, maybe you put Fable on that. But maybe you don't want to do that because Fable's gonna be really slow. Yeah.

42:59Unidentified speaker (speaker_2): So

42:59Unidentified speaker (speaker_0): you don't want to do that for everything. And so I think that, you know, if you think about these now for people who are actually deploying these, there's many more tools

43:08Unidentified speaker (speaker_2): that

43:08Unidentified speaker (speaker_0): they can do. And I think that's a very powerful mechanism.

43:11Unidentified speaker (speaker_2): Yeah, I think people also talk a little bit about um to your point. sort of uh public slash open world versus private, more closed loop systems and in the sort of public open world, like you have a robot in somebody's home or walking down the street with them or something else. You can't have like the lag of an internet connection or an API call because you're worried that the robot may freeze, fall over, something could happen. And so you want the intelligence to be embodied on the device as much as possible, or at least certain control mechanisms to provide. safety relative to the device. And that's the other aspect that people sometimes talk about in terms of uh state of the art models versus specialized fine-tuned models or just running on device and things like that, which is kind of interesting from a more safety perspective, I guess.

43:51Unidentified speaker (speaker_0): Yeah. I think I think that's a an important thing. I mean, it's a s there's a safety perspective and also there's just a speed.

43:56Unidentified speaker (speaker_2): Yeah.

43:57Unidentified speaker (speaker_0): Right. Um, that you you want, you know, your robotics typically to operate on a very fast clock cycle. Yeah. Um that is challenging to do with these larger models. Um, you know, historically that is typically solved by distilling into smaller models, whether that's something that the end user is doing, or whether you know, you have your Astra, and by the time that gets turned into a Luna model, you know, it can still work on some of the most popular use cases, but it's able to operate like very quickly and very efficiently. Actual and edge on device models, I think, are also, you know, likely to be very important here. Um and they're going to rely on prop typically they end up being solved by open source that, you know, then gets improved through one way or another coming from data from

44:41Unidentified speaker (UNKNOWN): the big labs.

44:43Unidentified speaker (speaker_1): So amazing run of Palantir and an OpenAI, OpenAI, you're used to doing things relatively fast, including with like difficult ecosystems. Uh it's conventional wisdom that working with large incumbents with physical world plants, trying to get them to change big processes is hard. Right. Like how do you think about the pace for Arda and what's possible here?

45:02Unidentified speaker (speaker_0): Yeah. So there is just a huge difference between being, you know, developing technology for yourself as a first party, like you know, building models for use in ChatGPT and, you know, building something for use in a large enterprise. And the the way that we're approaching this is that we're using the forward-deployed engineering model. And so I think, you know, sort of classically, uh, we developed this in Palantir, you know, way back in 2005. You know, the way I like to think about it is, you know, back in 2005, I spent a lot of time in secure spaces in Northern Virginia working with people with secret workflows that couldn't leave the building. And had to figure out how to build a product for them. Right. And, you know, turns out building a product for people that have secrets that they can't tell the world is actually really good training for building a product for any enterprise. Because any enterprise that works, like if they're working, they have secrets and they don't want to tell the world everything about their IP. And so the outcome of that process was the forward forward-deployed deploy. engineering model, which I think mostly Sean Senker, who's the CTO at Palantir, deserves the most credit for. But my job was to figure out how to build products for it. And so what that does is you basically take your engineers

46:15and you put them, you know, embed them with the customer. Um, I would not develop a feature unless I had a customer that was ready to use that feature. And so when I think about the product. roadmap, a lot of the the the beginning point for the product roadmap is not okay, I'm going to go build this feature. It's how do I find a customer that needs this feature so I can go build it. As a consequence, then you need to start embedding with understanding, you know, seeing these enterprises. You still want to move fast though. So how do you move fast? Well typically you want to start by finding a small number of power users.

46:51Unidentified speaker (UNKNOWN): Um,

46:52Unidentified speaker (speaker_0): you know, classically, uh, at the agency way back in the day, there was a woman named Sarah Adams that was like the only user of Palantir for about a year. Um, and we learned I would go visit and you know, I would meet her, and you know, she would be like, This stuff doesn't work. And I'd be like, Okay, I've got to go fix this. And eventually it worked. And then suddenly lots more people used it, right? But it's the power of that just very small number of very strong users that are willing to fight through the pain with you.

47:21Unidentified speaker (UNKNOWN): Um,

47:21Unidentified speaker (speaker_0): you know, I told the story about our first user for Arta, our first user at Androl of uh a sort of lead manufacturing engineer who was trying to do the virtual build for an actuator. Well, what I didn't say is that this was a tremendously frustrating process for him, being the first user, and he was able to make you know, tremendous improvements and be able to move very fast with that actuator. But it only works because our head of product manager was sitting right next to him. You know, and every time he did something, it didn't quite work the way he wanted it to. And, you know, our head of product manager was just sitting there like taking notes. And then two weeks later, all those things worked.

47:57Unidentified speaker (UNKNOWN): Thank

47:57Unidentified speaker (speaker_1): you, user number one.

47:58Unidentified speaker (speaker_0): Exactly. Exactly. And so, but but like it's not like this process ends. because you know, six months later, we're at Anderrol, they're like, okay, go see if we can figure out how to build missiles faster. Well, this only works if we find that power user who's on the missiles team, who's now interested in working on a product that has uh a hundred times more parts than the actuator, two orders of magnitudes more parts than the actuator that we originally worked. with. Yeah, once again, out of the box, the product doesn't quite work. You find that user, you you with that user, you fight through to make it work. And then they can introduce it to everybody else. And so you're the, you know, uh, you know, classically, you when you're doing enterprise Price sales, you want to find that visionary early on. You know, this is from crossing the chasm, the classic business book. Classically, that visionary is a sponsor, and you do need that sponsor. But what the book doesn't talk about is you actually also need a user, someone who's actually willing to use your software to go accomplish an outcome.

49:01Unidentified speaker (UNKNOWN): And

49:01Unidentified speaker (speaker_1): if you do both these things, then even very large organizations will work on their problems faster than you expect.

49:07Unidentified speaker (UNKNOWN): Well,

49:07Unidentified speaker (speaker_0): there's not really any such thing as Is a very large organization. Every, you know, it's people will be like, well, what does OpenAI want? It's like OpenAI doesn't want anything. Sam wants something, Greg wants something, like, you know, this person, you know, this person down here, this research scientist on this team wants something. And big organizations are just an assemblage of very small teams. And so, you know, you have to proceed problem by problem, team by team. You know, one of the really interesting ideas in the FDE approach is you're not just throwing the software over the wall to them. You are partnering with them to achieve. an outcome. For example, you know, figure out how to build missiles faster. And you are not successful unless the customer is successful. And once you achieve that outcome, you can then look around at all the other outcomes and you've earned the right to work on something bigger. In in our case, maybe that's we go work on an airplane, we go work on the Fury with Android, Andrew, right? But also it could be we work on an entirely different problem that's bigger in terms. of dollars for them. It moves the needle more for them. And maybe they wouldn't be willing to trust you, to trust your startup to solve that problem unless you had demonstrated some form of excellence. And so, you know, really within a lot of this within an enterprise is snowballing, you know, the The successes you've made, snowballing the outcomes you can achieve while still baking that into your product.

50:27Unidentified speaker (UNKNOWN): Cause

50:28Unidentified speaker (speaker_0): I think the the thing that a lot of people who try the FDE approach now get wrong is that they see this as okay, well, I'm just deploying some engineers at the site and they're gonna make the product work at the site. You know, so somewhat famously, like people didn't believe the FDE approach worked, right? Like it was widely mocked in you know, 2015 to 2020. Folks

50:45Unidentified speaker (speaker_1): thought there would never be leverage on.

50:46Unidentified speaker (UNKNOWN): the phone.

50:46Unidentified speaker (speaker_0): That's right. Yeah. These are just consultants, like, oh, Palantir just sends a bunch of consultants to do things.

50:51Unidentified speaker (UNKNOWN): And

50:52Unidentified speaker (speaker_0): then people started doing it when Palantir became a $300 billion company and started profit. And the difference is that they weren't just consultants, right? The FDE strategy is a product strategy that teaches you how to build products. The FDEs are the gap between the product and the user, or they're they're filling the gap between the product and the user. And if they have to keep filling the gap, you get no leverage. Instead, you have to productize what they're seeing. and bake that back into what you're doing. And so that's what allows you to move quickly and dis and and discover new things and bake those into your products.

51:23Unidentified speaker (UNKNOWN): So

51:23Unidentified speaker (speaker_2): if we take all this to the extreme and you have an army of FDEs over time who are deploying across multiple industries, you're learning very quickly, you're iterating, you're building a lot of product that then gets reused at scale. Where does this all go? Like what is your vision for the endpoint of all this?

51:37Unidentified speaker (speaker_0): Yeah. Well,

51:39Unidentified speaker (UNKNOWN): I

51:39Unidentified speaker (speaker_0): I I would go back to to think about superintelligence at the beginning. So we started by talking about, you know, what it looks like to have digital only superintelligence. And, you know, yes, the need for laying out lots of computers, that's a requirement for superintelligence. But superintelligence also has consequences. So, you know,

51:57Unidentified speaker (UNKNOWN): I

51:57Unidentified speaker (speaker_0): love Dario's phrase of, you know, a country of geniuses and a data center. But if all they can do is sit in that data center and think, and like you can send them emails and they can send you emails back, like they're not going to be able to do very much. They're kind of trapped in the data center. And so, okay, so intelligence has become cheap. What are the complements to intelligence? Well, perception on the one hand and action. As we have all these AGIs, you know, right now, humans are the eyes. and hands of the intelligence that's in the data center, right? And instead, what I think we're gonna be seeing over the next 10 years is this tremendous profusion of various different kinds of perception, various different kinds of manipulation. You know, there'll be you know drones and sidewalk robots, every physical thing that we own will become, you know, perception enabled. and hooked up to the cloud. And, you know, whatever it does will become controlled by AI. And then all

52:57sorts of things that we want to do, we will be doing because there are robots that are out there doing these things for you. Some people think that they will just be an army of humanoid robots. I think instead we're going to see just this tremendous profusion of different forms. We think that, you know, the world is made. is for humans. But in many ways, you know, if you if you look at sort of the broader world, it's actually, you know, we'll probably see robots that begin to look like all different kinds of animals, right? There will be robots that fly, there will be robots on four legs, there will be robots that carry burdens, there will be robots that uh, you know, plow fields. Um, I think we already have those, right? With tractors.

53:35Unidentified speaker (UNKNOWN): Um,

53:36Unidentified speaker (speaker_0): and so

53:38Unidentified speaker (UNKNOWN): the

53:39Unidentified speaker (speaker_0): the the part of the story of superintelligence is not just,

53:43Unidentified speaker (UNKNOWN): you know,

53:44Unidentified speaker (speaker_0): creating superintelligence and having it live in the data center and maybe scheme about how to take over the world. It's you wanted to actually be able to go out of the data center and help people and do things for us. And so, you know, that I think is the the call to action for manufacturing and why what we're doing, what me what m this you know manufacturing writ large is doing in this world of superintelligence.

54:08Unidentified speaker (UNKNOWN): You know,

54:08Unidentified speaker (speaker_2): von Neumann and I don't remember when he came up with this. He came up with uh with uh von Neumann probes, right? These self replicating probes that you send out of space, they land on an asteroid, they mine it, or they land on a planet, they mine it, they produce more versions of themselves and they keep spreading.

54:21Unidentified speaker (speaker_0): Yeah.

54:22Unidentified speaker (UNKNOWN): And

54:22Unidentified speaker (speaker_2): so it's the way to effectively explore the universe at scale in a sort of distributed system way where you have things self-assemble and then keep going. And it's exponential, right? You land, you make 10 copies, those 10 copies go off and make 10 more copies and you kind of spread through the universe.

54:35Unidentified speaker (UNKNOWN): Um

54:36Unidentified speaker (speaker_2): do you view that as an eventual manufacturing model on on our planet? In other words, eventually do you have these self-replicating factories or what what do you kind of view as the end state from an almost uh physical form perspective?

54:47Unidentified speaker (speaker_0): Let's let's like narrow in and think about this concretely, right? So you know, uh, you know, maybe you have a problem that you need to go build more bulldozers. And so you need to build a bulldozer factory, right? Well, right there you have a recursive loop where the

55:02Unidentified speaker (UNKNOWN): the

55:02Unidentified speaker (speaker_0): right way to do this is that you're deploying your intelligence to go manufacture more things. Um, and so, you know, I think I think very, very quickly we will see, you know, yes, uh factories that build other factories. But like that, that's machine tools, right? Like that's

55:17Unidentified speaker (UNKNOWN): what we have,

55:18Unidentified speaker (speaker_0): right? But the difference is doing this in an intelligent way, where you know, the it is a tremendous amount of human work to be able to go build a factory. It is a tremendous amount of human work to be able to design machine tools for that factory. All of that, I think, is it going to become increasingly automated. And and it will have to in order to be able to support the vast. network of things that we need in order to

55:41Unidentified speaker (speaker_2): it's basically the digital singularity versus the physical singularity.

55:44Unidentified speaker (UNKNOWN): Yeah. Or

55:45Unidentified speaker (speaker_0): and

55:45Unidentified speaker (UNKNOWN): not even

55:45Unidentified speaker (speaker_0): versus just alongside like part of the digital singularity involves converting increasing fractions of the Earth's crust into data centers.

55:53Unidentified speaker (speaker_2): Yeah.

55:53Unidentified speaker (speaker_0): Right. Um and maybe those data centers are going to be in space too.

55:57Unidentified speaker (UNKNOWN): Um

55:57Unidentified speaker (speaker_0): and like that is a physical process and it's going gonna to have to be directed by AI just. like increasing human civilization across the world was directed by human intelligence. So I think that's a that's a very near-term outcome. You know, that's that's like what we're doing in the next 10 years as a as humanity. And then in the longer run, I think,

56:17Unidentified speaker (UNKNOWN): well,

56:17Unidentified speaker (speaker_0): once you have self, you know, it it machines can self-replicate and they are a better form factor for spreading between. the stars than humanity is. And so it seems almost certain that it will happen. And so, you know, then you want to make sure that this is a thing that's happening that is that is liberating humanity, that you know, humanity is benefiting from, that humanity uh remains as much as possible in control of rather than something that, you know. where humanity is a little sideline that just remains here on Earth and watches a lot of uh videos, uh, which probably will also be generated by AI.

57:00Unidentified speaker (speaker_1): They're going out there to terraform for us so that we may also explore the stars.

57:05Unidentified speaker (speaker_0): Exactly. I think we will also have a role to play if we choose to take

57:10Unidentified speaker (UNKNOWN): it in

57:11Unidentified speaker (speaker_0): getting to the stars.

57:12Unidentified speaker (speaker_2): Why do you think um I mean this is back to the firm. Fermi paradox, right? To some extent, AI and superintelligence and AGI and these self-replicating factories, they all suggest we should have come across other intelligent life sooner, actually. Right. In other words, it increases the likelihood of running across it than if you just had, you know, our biological forms trying to make it through space on our own.

57:33Unidentified speaker (UNKNOWN): Actually,

57:33Unidentified speaker (speaker_0): I disagree. Oh

57:34Unidentified speaker (UNKNOWN): and So I think this

57:35Unidentified speaker (speaker_0): is the resolution to the Fermi paradox, uh-huh. Which is uh if you think about it, what is the most effective

57:42Unidentified speaker (UNKNOWN): uh you

57:43Unidentified speaker (speaker_0): know civilization spreading through space? It's one that you know goes to a star system, very quickly sets up a factory, builds a lot of self-replicating probes, sends those out. And so you have basically a wave of civilization threading spreading through space at the speed of light.

58:02Unidentified speaker (UNKNOWN): Yeah.

58:02Unidentified speaker (speaker_0): And so why don't we see that? Because it's spreading through space at the speed of light.

58:08Unidentified speaker (speaker_2): I see.

58:08Unidentified speaker (speaker_0): And, you know, uh, we will only see it, you know, a few minutes or or perhaps a few years before it arrives. Aaron

58:15Unidentified speaker (UNKNOWN): Powell

58:16Unidentified speaker (speaker_2): Isn't the argument that it should have arrived by now? In other words, there's arguments in terms of what's the density of potential civilizations throughout space and the rate at which they arise, and then therefore. if you send something out at the, you know, sub sub speed of light, what is the rate at which it would eventually get to you, et cetera?

58:33Unidentified speaker (UNKNOWN): And

58:33Unidentified speaker (speaker_2): so that would suggest sparseness in the universe, right? In terms of

58:37Unidentified speaker (UNKNOWN): well,

58:37Unidentified speaker (speaker_0): I think I mean, I think this is a mystery, right? Is like why are we the first ones?

58:41Unidentified speaker (speaker_2): Uh huh.

58:42Unidentified speaker (UNKNOWN): Um,

58:42Unidentified speaker (speaker_0): you know, and there's some evidence that says if you look at sort of the uncertainty of all the different factors

58:47Unidentified speaker (UNKNOWN): uh

58:47Unidentified speaker (speaker_0): in the Drake equation, that you end up seeing that, okay, well, maybe. actually it's quite unlikely. Uh-huh.

58:52Unidentified speaker (UNKNOWN): And

58:52Unidentified speaker (speaker_0): so it is plausible that we're the first

58:55Unidentified speaker (UNKNOWN): civilization.

58:55Unidentified speaker (speaker_2): Yeah, exactly.

58:56Unidentified speaker (speaker_0): Yeah. But I think

58:57Unidentified speaker (speaker_1): Or they're just very far away.

58:58Unidentified speaker (UNKNOWN): Yeah,

58:59Unidentified speaker (speaker_0): or that it's very far away. Right. Yeah. It's like somewhere out there in the universe, but like, you know, our light cones intersecting, you know, is gonna take some time. But I think, you know, to some extent, evidence shows that this hasn't happened, right? And then, you know, the question of are we going to run across,

59:17Unidentified speaker (speaker_2): you

59:17Unidentified speaker (speaker_0): know, alien civilizations? I think this argument says, no, we're not. We're only going to attack them, you know, a little bit before they arrive. Or perhaps we will be the wave. Hopefully, we will be the wave of civilization spreading out through the galaxy at light speed.

59:33Unidentified speaker (speaker_2): Amazing. Thank you for joining us today.

59:35Unidentified speaker (speaker_0): Thanks for having me. This was

59:36Unidentified speaker (UNKNOWN): fun.

59:39Unidentified speaker (speaker_1): Find us on Twitter at NoPriors Pod. Subscribe to our YouTube channel if you want to see our faces. Follow the show on Apple Podcasts, Spotify, or wherever you listen. That way you get a new episode every week. And sign up for emails or find transcripts for every episode at no dash priors.com.

This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.
All insights