正在加载视频...
视频加载失败
Please enjoy this Cheeky Pint / Dwarkesh Patel crossover with Elon Musk. Dwarkesh was most interested in how Elon is going to make space datacenters work. I was most interested in Elon's method for attacking hard technical problems, and why it hasn’t been replicated as much as you might... show more
35 条评论

@dwarkesh_sp @elonmusk “People are too afraid to try things. They shouldn't be afraid of failing. They should just go and do it.” — Elon Musk How it started How it going

Spotify: Apple Podcasts: YouTube: Substack:

What stood out isn’t space GPUs or Optimus in isolation.. it’s the method. Elon doesn’t solve problems linearly. He collapses them to first principles, removes hidden assumptions, then rebuilds the system as if it never existed. That’s why it’s hard to replicate. Most people optimize inside constraints. He questions why the constraints exist at all. Space datacenters, xAI, Optimus — same pattern: physics → cost curves → iteration speed → scale. The takeaway isn’t the tech. It’s the operating system for thinking.

@dwarkesh_sp @elonmusk Hope you enjoyed our bar @neuralink 🙂

@dwarkesh_sp @elonmusk one word timestamps go hard

@dwarkesh_sp @elonmusk so glad there's a conversation where elon isn't being asked the same questions for an hour long panel with an arbitrary time limit

@dwarkesh_sp @elonmusk In my next life I hope to be able to visit Mars and maybe spend time living on Mars. People will wonder what it was like to live in the era of Elon Musk and President Trump.

@dwarkesh_sp @elonmusk “Data centres in space…”

@dwarkesh_sp @elonmusk It's always sunny in space

@dwarkesh_sp @elonmusk Everyday I’m so impressed with Elon’s range of knowledge. He understands so much at a level of detail of people that work on the problems do. Most CEOs would be lost if you put them to these detailed questions.

@dwarkesh_sp @elonmusk Thank you guys. Amazing work. We can’t have enough of Elon Musk

This was a great crossover—basically a three-hour deep dive into the future. 🚀 The most fascinating part was how Elon approaches hard technical problems from first principles, which explains why his methods are so difficult to replicate. Space GPUs, Optimus, xAI, alignment, China, management—everything connected into one bigger picture of where technology and power are headed.

@dwarkesh_sp @elonmusk “It’s harder to scale on the ground than to scale in space” What?

@dwarkesh_sp @elonmusk That!

these two lads were like possums in the headlights chatting with Elon....intellectual lightweights vs one of our World's heavyweights the hosts spent so much time beating to death their own pov on certain topics, they tried endlessly to get Elon to agree with them, they just needed to acknowledge his clearly and politely stated differing pov and move on....but their own ego's got in their way and listening to them rant on n on at times was so damn boring.... Look at the DOGE stuff as an example (and there were other examples), these two hosts were so condescending, neither of them have had insight into the workings of Govt like Elon has, he's been into the belly of that beast, he setup DOGE and discovered the fraud, waste and inefficiencies which DOGE continues to find today..... but these two fools were essentially calling Elon a liar when he stated how much fraud is/was occurring within Govt... and if that wasn't embarrassing enough they topped it off by spewing a dipshit comment about the Govt fraud that went something like "well why didn't you just stop paying it."🤣🤡 Honestly these two kids need to get back to their gaming and drinking, leave these heavyweight interviews to more capable and mature hosts my scores for interview... Hosts: 1/10 Guest: 100% for remaining as Calm as a Meditating Monk

@dwarkesh_sp @elonmusk Alignment is THE issue of our time. An AI that lies will do anything, and we don’t want that future. Maximum truth seeking is existential to human flourishing.

@dwarkesh_sp @elonmusk WOW - this is huge! Cheeky Pint is definitely one of my most favourite podcasts and the episode that you did with Dave Ricks - the CEO of Eli Lilly $LLY was a true masterpiece 🧵👇! Kudos for an excellent job and Godspeed!

@dwarkesh_sp @elonmusk 3 hours of elon and the section i'll skip to first is "Management" at 01:40:46 the man runs mass 6 companies. either he's cracked time management or he's 4 people in a trenchcoat.

@dwarkesh_sp @elonmusk So much going on in Elon's mind.

@dwarkesh_sp @elonmusk

The United States cannot compete with China on human labor alone because of China’s fourfold population advantage and an industrial capacity that is already three times larger. Elon Musk sees humanoid robots like Optimus as a recursive force that can solve this by manufacturing more units of themselves, potentially expanding the economy by 100,000 times. This requires designing every component from physics first principles, especially the robot's hand, which is more complex to build than all other hardware parts combined. Moving toward a pure AI and robotics corporate model is likely the only path to maintaining industrial relevance in a global market. Episode notes here if you prefer reading

I'm watching using @Starlink while Grok is open on X, and all of the little stars are running around the screen listening to Elon talk about data centers in space... It's totally going to happen. In fact, it's more fun to think of it this way, so that makes it all the more plausible because the future is fun.

@dwarkesh_sp @elonmusk What are these? Pints for ants?

@dwarkesh_sp @elonmusk I feel I could listen to this daily for a few months to get a better grasp of what's to come. I also feel like we should be preparing for a change in our timeline that is rapidly approaching. A wild unfolding of reality.

@dwarkesh_sp @elonmusk YT Version:

@dwarkesh_sp @elonmusk Space datacenters powered by constant solar + beaming the compute back? That’s wild. 🔥 If the economics and latency ever work out, goodbye earthly power constraints. Let’s goooo✨💥

I’m literally at the part using customer service as an example of revenue using AI. I DONT WANT TO SPEAK TO AI. I want to speak with a HUMAN, and not be charged for it. There is already too much navigation just to get to a human now. AI cannot answer my question the way I need it to be answered. I need the human experience of conversation

The biggest takeaway is that @elonmusk understands the difference between what sounds easier on paper (build more turbines on earth) but in reality contains more bottlenecks than what sounds more difficult on paper (build outs in space) but is actually easier in practice. And that itself is a superpower.

Review of the interview with particular focus on What's New & Reading Between The Lines. What's New - Elon dropped a bombshell on space-based AI compute - predicting it'll be the cheapest and most scalable option in roughly 30 months. Endless solar power in orbit plus Starship driving launch costs way down means orbital data centers could hit hundreds of gigawatts per year within five years. SpaceX and Tesla are already ramping solar production toward 100 gigawatts annually to feed this vision. 🚀 - Optimus humanoid robot progress sounds massive - Gen 3 aiming for roughly a million units per year, Gen 4 targeting 10 million. They're setting up an "Optimus Academy" with tens of thousands of real robots running self-play training, backed by millions more in simulation. This recursive loop could explode manufacturing capability faster than anyone expects. - xAI's trajectory looks game-changing - planning to emulate digital humans by year-end, which unlocks trillion-dollar opportunities like customer service at scale. Longer term, Elon sees pure AI and robotics companies completely dominating global economies. He also hinted at TeraFabs that could churn out millions of wafers monthly to crush chip supply bottlenecks. - On the government side via DOGE, he called out hundreds of billions lost to fraud and waste - with straightforward fixes that could save enormous sums. But the real hurdle is bureaucratic inertia in what he calls the biggest "corporation" with a monopoly on violence. Reading Between the Lines - Elon's relentless focus on limiting factors ties everything together - energy is the bottleneck today, chips tomorrow, then recursive robot manufacturing. His unusually high pain tolerance explains bold pivots (like switching Starship to steel) that others shy away from due to risk aversion. He only dives deep into details when they're the true constraint. - Quiet but stark warning on China - without breakthroughs like Optimus humanoid robots, they'll dominate manufacturing and AI thanks to sheer population scale, work ethic, and refining capacity. The US counterplay? Flood the labor gap with robots - enabling way more ore refineries and economic leverage. - AI alignment came across thoughtful and realistic - humans won't stay in control of superintelligent systems indefinitely, but embedding truth-seeking values (no deception, reality as the ultimate verifier) offers a shot at bright outcomes. He's cautiously optimistic that this path propagates consciousness - ours included - toward vastly better quality of life. 🤔 Summary - This three-hour pub conversation with John Collison and Dwarkesh Patel was remarkably casual yet absolutely loaded with insight. Elon unpacked why space data centers solve Earth's power crunch for AI scaling - unlimited solar and frequent cheap launches will soon outpace anything ground-based. - He laid out xAI's truth-seeking approach to alignment, the coming wave of digital coworkers, and physical Optimus humanoid robots as economic disruptors. Management lessons emphasised urgency, prioritising execution over credentials, and attacking bottlenecks directly. - China earned genuine respect for manufacturing strength, while US competitiveness hinges on robotics recursion. DOGE frustrations highlighted government's sluggishness, yet the overall tone stayed hopeful - if we move fast enough, space compute, robot legions, and truth-seeking AI can extend civilisation dramatically. - Wildly optimistic vibe throughout. The future feels electric if we keep pushing. What a marathon interview - so much to chew on. 😊

@dwarkesh_sp @elonmusk I am a power plant operator, been in the industry for 17 years. Data center demand is supposed to triple the total demand in the phoenix area in the next 5 years... its crazy...

@dwarkesh_sp @elonmusk Interviewers should avoid cutting Elon off too much...

@dwarkesh_sp @elonmusk

@dwarkesh_sp @elonmusk I get the energy side of it, but I'm unsure about immortal hardware in space, as I come across many hardware failures in data centers. How would one service and replace them?

@dwarkesh_sp @elonmusk "Don't take my word for it, read the GAO report" Problem: Elon never actually read or understood said report

@dwarkesh_sp @elonmusk The value of this three hours compared to Elon's net worth is literally 1.1B...insane to consider.


