Elon Enters the Chip Race, the S&P 500 Repricing, and Human Drivers Will Become Illegal | EP #242
In a Nutshell
Elon Musk's TerraFab announcement targets 1 terawatt of AI compute annually (50x global capacity) via a massive Austin fab for Tesla, xAI, and SpaceX chips, scaling to petawatt orbital compute with lunar mass drivers, potentially valuing his empire at $100 trillion amid chip shortages delaying self-driving cars. Autonomous vehicles and eVTOLs like Joby's will slash crashes 90%+, collapse car ownership 50-90%, eliminate parking, and reprice real estate toward abundance, with human driving likely banned in cities within 3-5 years. AI disrupts jobs (25% automation, token-tracking metrics), compresses S&P valuations to 2-7x FCF, accelerates space race via Starship, and advances cryonics, model distillation, and self-designing chips for exponential abundance.
These notes were generated by AI and may contain inaccuracies.
Elon's announcement of the TerraFab is the number one story this week. > "This is the most important endeavor in human history by far. In order to understand the universe, you must explore the universe." He's building a galactic factory. Current global AI compute output is 20 gigawatts. TerraFab aims for 1 terawatt of AI compute per year across Tesla, xAI, and SpaceX—50 times current global production. Building two kinds of chips: edge inference chip for robots and cars, and high-power rad-hard chips for space Dyson sphere. Fab in Austin, eventually 100 million square feet. Near-term: 1 terawatt in single-digit years. Long-term: petawatt from lunar mass drivers.
Self-driving car uses a full GPU. > "When is it going to become illegal for humans to drive?" Technology and demand will be ready before chips. Figure out more compute with less silicon for this use case to become an instant billionaire. Heading toward a $100 trillion company, potentially the largest on and off the planet.
Welcome to Moonshots, episode of WTF. Hosts: Peter (likely Peter Diamandis), DB2 in Boston, AWG at home base without saucer-separated Enterprise 1701D Lego set, Dr. Exo Salem (Sem) at JFK. Sem flying to Brazil for 40 hours. Format change: deeper conversations on limited subjects, covering breaking news. Mission: get listeners future-ready, excited about abundance, opportunities for entrepreneurs, investors, students, parents. Focus on supersonic tsunami of most important tech. Aim to be #1 podcast on AI and exponential tech. News story count exploding exponentially.
TerraFab puts all parts of Elon's Lego puzzle together for massive capabilities. Elon approached chip manufacturers: > "I will pay you for as much production rate as you'll give me." None move at Elon speed, so he's building his own fab, repeating playbook from launch and EV/autonomous industries.
Dave: Most important endeavor unlocks everything. Scale at humanity's need, but physical constraints: ASML EUV machines, complicated supply chain. Estimated 50x current chip productivity last podcast, confirmed accurate. Elon alluded last summer. Cover story: chips hardened for cars and space, not competing directly. > "I will buy everything Samsung can offer me, but you're not offering enough. So I will still build all these chips and buy everything you give me." Austin fab: full vertical integration under one roof for rapid chip design iterations. TSMC conservative; Elon planned 50x bigger.
Target: 1 terawatt compute per year in orbit. Requires 10 million tons mass to orbit per year. Next-gen Starlink satellite: 1 ton each. Needs 274 launches per day on Starship—a launch every 5.3 minutes. Comparable to airline business.
If TerraFab ramps in next 5 years, tremendous geopolitical implications: could accelerate or mitigate WWIII and Chinese invasion of Taiwan. Lunar petawatt compute: 3.1 million tons lunar mass. Scaling to exawatt: 3% of moon's mass. > "The moon had it coming." Slated for disassembly to build Dyson swarm. Joint Tesla (20%), xAI, SpaceX (80%)—cornerstone for grand unification of Elon's projects.
Heading to $100 trillion company. Terrafab worth multi-trillions. Needs $25 billion initially for one fab; 50x world production requires 50x that ($1.25 trillion total capex estimate). TSMC valued at $1.7 trillion; TerraFab expected 70% of TSMC output. Plus Nvidia ($4.5 trillion). Real value: TSMC + Nvidia x 50. SpaceX IPO prediction markets: from $1.5T to $2T+, pricing in Terrafab.
Patron saint of exponentials. Aims vertical future, flat path. Redefines game at bottlenecks. Launch cadence forces operating model change. Elon's timing predictions 15-20% accurate, but direction matters. > "Shoot for the stars and if you get to the moon, who the hell cares?"
1 terawatt aspirational, 5 years timeline (50x). All chips, power, data centers pay off—no cannibalization. Need all compute. Self-driving may get deprioritized: full GPU for driving vs. brain surgery, new math/physics. Terrestrial critical for national security, backup if space fails. All process nodes (3nm, 5nm, 2nm, 1.6nm) run full throttle.
In production hell next 5 years, expect new material physics, process engineering. Alternatives to photolithography. Superintelligence designs systems. Need humanoid robots to build fabs—no workers. Secondary inventions massively beneficial, like carbon fiber from space.
Laying down single atoms via self-organizing process. AI discovers non-lithography future in 1-2 years, manufactured in 5-7 years.
Claude estimate: initial $20-25B capex, full buildout $150B-$500B. Annual cost savings for captive chips (Optimus, Cybercabs), external revenue. Implied enterprise value: $1T+ to multiple trillions. Numbers low; 50x world AI chips = category operator, beyond car company (TSMC + Nvidia).
No monopoly worries: MTP (massive transformative purpose) beneficial. Competitors emerge (e.g., Brendan Foody type). Antitrust needs competition; Elon invites it. Multiple Dyson swarms. Google: $300B revenue, $100B+ cash flow, own chips, lacks rockets (Eric Schmidt with Relativity Space). Abundance story: massive AI compute increase.
Elon thinks two moves ahead (now 6 years vs. 20). Skips fission: solar in space after earth solar (no turbines/reactors). Dyson sphere now forefront; timelines shifted, before fission licensing.
Plans grandiose, timed pre-IPO to excite. Still love audacity. > "If you shoot for Mars and end up at moon, you're still way up."
Here is our second conversation story about the future of human transportation. Robots are getting their driver's licenses, flying cars are taking flight. Waymo hit 170 million fully autonomous miles, equivalent to 200 human lifetimes of driving with 92% fewer serious crashes. They have 3,000 vehicles in 10 cities. Uber invested $1.25 billion in Rivian with plans to deploy 50,000 fully autonomous robo-taxis. Waymo data shows extraordinary crash reduction versus human drivers. Tesla's Full Self-Driving and Cybercab show similar incredible data.
Joby Aviation, started by JoeBen Bevirt over a decade ago after selling Velocity 11, partnered with Rob Nail. Joby flew over the Golden Gate Bridge in San Francisco. It's an eVTOL (electric vertical takeoff and landing), called flying cars. Joby is testing its first FAA-conforming aircraft, demonstrating reliable design repeatability, with demonstration flights over the Golden Gate Bridge. Joby and Uber announced Uber Air powered by Joby; Uber Elevate was sold to Joby. Competitor Archer Aviation has the Midnight aircraft, first to achieve 100% FAA acceptance of its eVTOL means of compliance. Flying cars will operate in the US in the next 18 months, in LA in a big way by 2028.
When will it become illegal for humans to drive? It will start with city centers and broaden out. Flying cars enable full urban redesign, making land abundant. Real estate becomes abundant; flying across the US shows it's empty. Between Toronto and Chicago airports, there are 10,000 islands and lakes—no scarcity problem, but mobility and accessibility issues. Compress travel time, reprice real estate. Shoutout to JoeBen Bevirt for building hardware over a decade against regulatory and market hurdles—Nobel Prize in patience.
eVTOLs will move quickly as they don't risk crashing into houses like road self-driving cars; they will be autonomous from birth. First airplanes will be piloted: single pilot, four passengers, rapid recharge at vertiports, average flight under 10 km (e.g., Santa Monica to Dodger Stadium avoiding the 10). Autonomy comes with data and demonstrations. FAA regulations require piloted starts; AI command and control is being worked on for cars first. Middle East already doing self-flying versions; short 2-year window. Human driving bans will happen quickly, like indoor smoking or drunk driving. Tipping point when voters say humans put children at risk. Self-driving is 90% safer, soon 95-97%. Human car crashes cause unbelievable tragedy; number one cause of death for under-5-year-olds in first world, devastating families. TV ad campaigns in 3-5 years with ugly images will make public road driving inconceivable—go to test tracks or country roads. Chip shortages may delay despite technology and demand. Solve more compute with less silicon for this use case to become billionaire.
"Friends don't let friends drive drunk." → "Friends don't let friends drive." Period.
Alex: Never illegal; redefine driving to higher abstraction levels. With FSD 14, tell it where to go, converse with Grok for minute steers. Human-on-the-loop AI driver hybrid keeps safety with human feel. Future: push car as fast as wanted, car knows limits (traction, road surface), prevents stupidity 1% of time. Accelerator pedal becomes "Mad Max" mode.
With Uber partnering Waymo and others, will Cybercab compete? Yes, very competitive market driving abundance and UHI. Dozen companies compete on quality, price, bringing costs down. Cybercab priced at ~$30K, people can buy. Goal: buy 25-50 in Santa Monica as fleet owner generating revenue.
Complete collapse in car market structure: 100 million new cars/year sit empty 94% time. Drop by 50% utility, collapse half the industry. Cars last near-infinitely (e.g., 2017 Tesla Model 3 to 1 million miles). Reshuffles car services, maintenance. Uber SFO to San Fran ~$200 mostly driver cost; remove driver, 10x cheaper (10-30 cents/mile, 4-5x cheaper than owning). Shrink vehicles needed 5-10x.
Profound real estate impact. Autonomous vehicles reduce vehicles 10x, no parking needed. AI orders vehicle automatically as you approach door. Convert garages: three-car garage became extra bedroom, storage, workout gym. Personal garages go away; rethink space. Most don't like driving—no one asks Uber driver to let them drive.
Actionable advice: If young, city job, living in city (60% listeners), keep renting, buy second home in beautiful hard-to-reach spot for later. By 2034-2036 with 10x wealth, those spots coveted. Buy near airport in city. Island real estate 10-100x more accessible, values up. Downtown LA: 30% blacktop parking, 60% land area parking in Los Angeles—becomes gardens, greenland, parks. Stadium parking lots become usable space; rethink tailgating, business opportunities.
Second/third-order: Repurpose parking garages. Respread of suburbia—invest less in urban centers with AV "virtual subway" anywhere-to-anywhere. Extrapolate: autonomous "whenevers" (office buildings or homes as AVs). Live in AV social network; syncs for meetings (Boston morning, DC evening, Chicago next day). Humans become internet packets routed by autonomous system.
eVTOL limited by local landing; need hub-and-spoke vertiports within 5-min drive, glued by Uber platform: AV to eVTOL site (10-20 km), then AV. Hyperloop: joined early company, Virgin involved, raised ~$100M, didn't proceed due to material science. Benefits supersonic point-to-point (LA-SF, LA-Vegas). Which first NY-LA: Hyperloop or rocket (Starship point-to-point)? Rockets tough due to orbital velocities, re-entry over cities; offshore landing 1 km out. Rockets viable: Elon behind, vehicle exists, launches every 0.5-3 min. Hyperloop for commercial/container (lower g-forces/safety). Plane travel now productive with internet.
Goldman: AI could automate 25% US work hours (low estimate). PwC to partners: "If you resist AI, you have no place here. AI tool yourself or get out." G42 job listing exclusively for AI agents. Companies track employee AI token usage; Jensen: "$500K engineer not consuming 250K tokens deeply alarming." Like De Beers "three months' salary for diamond." Jensen cutting engineer salaries in half via tokens. Perplexity AI won appeal to run shopping agents on Amazon.
Goldman chart: AI automates ~25% work across categories.
PwC "adapt or die" doesn't go far enough. Run typical company with 20-25% employees as workflows go human-to-human → agent-to-agent. No job perturbation: create 4-5x more companies; big firms transition slower than startups/mid-market. Consulting partners may have no place—AI agents do strategy; need for implementation if better agents.
Hard for organizations to self-disrupt (hourly → outcome pricing). Replacements from other firms, not self-disruption. G42 listing real for AI agents (marketing job); close to reverse discrimination, "humans need not apply." PwC: non-AI users out.
PWC partners already enforce AI usage: if you don't use AI, get out. This is heading to AI-only jobs where humans need not apply. Most large companies are walking dead; their business models will be fundamentally disrupted in 2-5 years. The challenge is self-disruption before others do it, which is nearly impossible. Strategy: invite super talented young entrepreneurs to pitch how they would disrupt the company if funded, then fund the best to build an adjacent disruptor company.
Ideal, the company design firm, realized their methodology would leak, so they picked a crazy partner to build the disruptor: an open IDEO marketplace of design ideas.
Private equity exploits AI automation. Companies with great cash flow, even if doomed, see profitability soar in transition because AI does jobs for 10% (soon 2%) of human cost, no labor laws, overhead, or insurance. Private equity buys them, AI-ifies for more cash flow, then uses it for incubators, attracting entrepreneurs, or acquiring startups. Both Anthropic and OpenAI partner with private equity to buy and AI-enable companies. A new $100 billion fund announced for this.
Companies track individual employee AI token usage and set minimums per employee. Implemented targets across companies: 80% token cost, 20% salary. Huge job disruption in 2-3 years, then recovery by 2030-2032. Goal: be in the 20% humans when costs shift to 80/20 tokens/humans. Employees improving AI efficiency by 1% are worth more than cost-cutting. Race to 80/20 from transitory 2/3 token, 1/3 human or 50/50.
Tokens aren't perfect (can waste them), but use another AI to evaluate if tokens were valuable or vacuous. Jensen Huang quote: if a half-million dollar engineer doesn't spend at least $250,000 on inputs flowing to Nvidia, alarmed. Frontier Labs have token maxing competitions on leaderboards.
Tokens are the first measurable unit of productivity: introspectable, legible, analyzable unlike hours (naive input measure). Advice for CEOs: measure token spend (Jensen's metric is perfect). Gather all prompt history; AI can analyze efficiency, evaluate direct reports' prompts/outputs to cut/train bottom 20%. On Amazon Bedrock, prompt history auto-goes to S3 buckets. Grab data now before personal accounts; do not reimburse unseen AI. Use company infrastructure; feed to another AI (e.g., Claude 3.5 Sonnet).
Fountain Life uses AI and tech for long, healthy life. 3.3% of seemingly healthy members have undetected cancer. Early detection vastly improves cure rates. Full body MRI and early cancer detection screening not covered by insurance yet, but goal to democratize via research. Cancers often asymptomatic until stage 3-4. Visit fountainlife.com/peter.
Chamath article: AI compresses equity valuations to 2-7x free cash flow from S&P average 22x. AI makes disruption cheap/fast; no projecting FCF beyond 5 years. Breaks SaaS projections, investment paradigms, VC. S&P 500: $58 trillion at 22x; at 7x loses 2/3 value; at 2x loses 90%. S&P PE ratios deviating since 1950.
Chamath: Modern capital markets assume competitive advantages compound (moats persist, brands endure, network effects defend). Strip that: dismantle century of capital allocation.
S&P won't collapse. Pre-computer, ambitions were enduring oil/manufacturing; now dead. 22x assumes same product 22 years (Apple doesn't). Tailwind: 10x economy next 10 years, massive new wealth. Value management agility to roll with innovations (e.g., Elon). Massive S&P shuffling: huge winners, doomed incumbents resting on laurels (insurance, oil).
Provocative but flawed: like Ray Kurzweil's singularity firewall. FCF goes somewhere (infrastructure, lunar mining); capital expands explosively with abundance. Not compression; disruption narrow to software/SaaS. Broader: energy, farmland, infrastructure abundant; post-moat world better. Value agility, rapid innovation over predictable cash flow/seat counts.
EXO index: Top 10 Fortune 100 (flexible/adaptable/purpose-driven) outperform bottom 10 by 40x shareholder returns over 7 years. Capital flows to infrastructure/adaptive platforms. Only moat: living system learning faster (inner loop). All moats attacked: IP copied, switching costs shrink, scale weakens. If FCF visibility <5 years, rewrite public markets: reward renewables/optionality over scale/stability. Physical assets matter more (atoms > bits). SaaS model broken. Mid-market non-S&P already at 7x; S&P propped by passive index funds/401ks (Elon: don't do blindly). Mid/small caps get AI tailwind, tripling FCF as multiples drop; serious bargains.
Shareholder calls shift to rapid iteration/reinvention. Update EXO index for agents/communities of agents; breaks old model. New criteria: intelligence stack, MTP architecture, trust framework. Organizations become API calls (legal, etc.); firm = liability/purpose container.
Private equity (e.g., Advent) retools non-AI cash cows for AI, tripling FCF using agents 1/1000th time. Rip through assets (regulatory, data) with Gemini/Claude/OpenAI agents. Transformation wave like Star Trek Genesis project. Applies to public equity too: AI activist investors writing open letters to disrupt firms. Board members give CEOs top cover to reinvent.
NASA picks SpaceX for moon via Artemis; Starship takes astronauts more safely/economically. Boeing's Artemis 2 (SLS) launches April 1st post March 12 readiness review: Apollo 8-style circumlunar orbit (no landing).
New NASA admin (pod guest soon) elevates SpaceX. China intends moon landing by 2030 (echoing 1961 Kennedy). Potatoes grow in lunar dust (Martian); Project Hail Mary movie out. Asteroid Ryugu: all 5 nucleic bases (A,T,C,G,U) found, strengthening panspermia (life originated elsewhere, rained on Earth).
NASA SLS expendable (parts in every congressional district) vs. reusable Starship. Starship >2x delivered mass to orbit vs. SLS (Saturn V comparison).
Starship delivers twice as much mass to trans lunar injection (TLI) compared to SLS with Artemis. Launch costs and mission costs are key: SLS is expensive like the space shuttle, costing $1 billion per launch at four launches a year or $4 billion at one launch a year, due to the standing army of 20,000 humans operating it. Starship will do a clean sweep. Blue Origin is also in play.
Alex: Bets on how long before United Launch Alliance (prime contractor for SLS) gets acquired by Jeff Bezos or someone else for contracts and expertise. SLS seen as a jobs program or way to keep defense/intelligence capabilities alive. Starship is the obvious leader in a second space race. Humans landing on the moon in 2-3 years, then Mars, playing out as foreseen but decades late.
Building covered wagons, railroads, wagon train to the stars (Gene Roddenberry). Starship is extremely economical. Shift from government space theater to commercial space evolution. Exciting: nuclear bases (adenine, guanine, cytosine, thymine, uracil) found on Ryugu, taking life from scarcity to abundance. Expect microbial life ubiquitous on Mars, Europa, planets/moons.
Peter Jared (NASA administrator) predicts >90% probability NASA will imminently find evidence of microbial life on Mars—a sea change from prior ambiguous positions (frozen water, liquid water, hoped-for signs). Questions: How related to Earth life? Mars cooled first (~1 billion years earlier), life may have evolved there first. Martian meteorites reached Earth via asteroid impacts—did they carry life? Common genes possible. Independent life on Europa would be amazing.
Scenario: Life started on Mars or farther out, transported by asteroids. Bounded to solar system or propagates through deep space? Interstellar medium contains building blocks of life. Galaxy well-mixed on billion-year scale. Panspermia at galactic scale possible due to stellar velocities, close passes. Several generations of stars exploding and reforming with nebular mixing.
Genetic complexity extrapolation: Last common ancestor timing suggests first base pair ~1 billion years before life observed on Earth.
Drake equation adds up to 100%. Non-carbon-based life would be exciting. Dinosaurs extinguished by meteor; base pairs propagate via asteroids/meteors, but life often extinguished until cooling/settling allows evolution to intelligence—a system dynamic settling problem.
AI frontier fracturing into stealth arms race; anonymity new moat. OpenAI launches GPT 5.4 mini/nano: twice as fast, approaches full GPT 5.4 on coding benchmarks—models getting smaller/faster.
Mystery: Hunter Alpha (1 trillion parameter model) on OpenRouter—no attribution, 1 million token context, free, processed 160B+ tokens. Thought to be DeepSeek V4, but Xiaomi's AI team; stock up 5.8%. Xiaomi evolved from mobile phones (met young founders) to EVs to killer model.
Proliferation hard to contain: Prior models roadmap next ones. Trillion-parameter models will propagate worldwide with $50-100M investment, dropping with algorithmic improvements.
Distillation: Launch big model, then mini/nano using synthetic data from larger model to train smaller, faster, cheaper student model (OpenAI, Anthropic, DeepMind do this). Continues to work despite iterated amplification/distillation cycles—borderline magic. Suggests endgame: distilled black hole/neutron star model, maybe 1KB file with master equation for superintelligence on your phone.
Example: Distilled model on iPhone in airplane mode answers every question—knowledge of humanity in teddy bear, Thomas train set.
Moats shifting: No longer brand/capitalization. Data is great moat (proprietary for brain surgery, materials science, chemistry). Synthetic data possible but all boats rise. Advice: Replay Alex's points, generalize with AI, read docs—build distilled focused model solving problem better than anyone (instant business). Use OpenClaw on podcast episodes.
Sam Altman might cite: Data centers (Stargate pivoting to renting), best researchers (hemorrhaging, commoditized), billion+ users (enterprises more valuable; cheap Chinese open-weight models fine-tuned with Claude outputs compete). Baseline models commoditizing; value migrates to OpenClaw/higher frameworks.
Design Conductor by Veor autonomously built 1.5 GHz Linux-capable RISC-V CPU from concept to tape-out in 12 hours (vs 90 days traditional). Recursive self-improvement breaking out of software loop into EDA, then data centers/energy/economy.
Cynical view: RISC-V has unit tests for verifiable rewards/RL. But remarkable: AI designing own chips. Extends to robots building robots.
Impact: Senior design engineer $400K, team millions/year → lunch break. 50,000 hardware engineers? Unlocks use-case-specific chips (10x+ efficient). $23T chip/data center spend → hundreds of billions saved. Fabs handle daily mask changes. All engineers useful for diverse designs.
DOE $300M for Genesis project: AI across 20 national challenges (manufacturing, biotech, energy) using national labs data.
Good US industrial policy articulating grand challenges (e.g., fusion). First from DOE. Worked with Daario (DOE lead) at MIT.
Concern: China deploys hundreds of billions state-directed; this small but hopefully first. US less distorting, deeper private capital—better odds.
Rural Ohioans seek constitutional amendment banning data centers >25 MW—grassroots NIMBY. Drives to orbit (YIMOP: yes in my orbital plane). $10B per GW invested power; tiny land footprint, astronomical wealth. Insane to ban via amendment.
There's got to be a much better way to make a win-win than to ban something that's obviously going to benefit your state tremendously. You're in California. Alex and I are in Massachusetts. The way we make decisions through legislators is so messed up. Something like this could even get proposed is ludicrous. That's what really needs to change because when you talk to the governors, they're like, I don't want this. We're a representative democracy. They're supposed to be very, very smart people thinking about complex issues and then deciding what happens. You don't throw things like this out to a referendum of people who just got laid off. People saying, my access to clean water and energy and my consumer price index of energy is going through the roof. There are other ways to deal with this than banning it by constitutional amendment. That's insane. I've never seen a data center that affected the water supply. Data centers need a fixed amount of water to cool themselves. It doesn't drink the water. It just goes around in a circle.
Nvidia won approval to sell its H200 chips, its most advanced chips, to Beijing. That's a big deal. The realization is the ban didn't work. China was getting access to chips through third parties and China was developing its own competitive equivalent. This cost Nvidia tens of billions of dollars. Since it's not working and it's stimulating a homegrown equivalent of Nvidia in China, they said let's reverse policy. Question is is it too late. Everything that you just said is correct except the one part where when Jensen complains he lost tens of billions of dollars, every single thing he's manufactured is sold out for years to come. The fact that it didn't go to China, it definitely sold. Even if it was a dysfunctional Chinese design, it still got sold. Everyone in the world wants these things. So, he didn't actually lose any money. The embargo or the ban didn't work. China is doing its own thing now. But if you say, well, let's start selling them again. Maybe they'll stop. Not that that's going to happen. You cut them off. They're not going to forget. If I'm Beijing, variety of Chinese frontier AI labs are all slurping up as many H200s as they can get. Blackwells are now the frontier. Beijing is being kept a half step or two behind the frontier chips available to the US AI labs. If I were the Chinese Communist Party I'd be doing the moral equivalent of having people taste my water at this point in terms of these chips. Nvidia's been very public about countermeasures to prevent the wrong chips from ending up in the wrong locations. The Chinese government is suspicious at the circuit level of American chips. They're looking now at our chips with renewed scrutiny to see what else is in these chips that we're shipping to them, what algorithms are embedded deep inside.
Scientists successfully froze an entire pig brain while locking in the cellular activity with minimal damage. This is cryogenics in a large mammal. The pig has organs, heart, liver, lung, kidney, and brain on the order of human organs. This is significant. This is from a company I informally advise named Nectome. I have another company with the founder of Nectome involved, Eon, focusing on whole brain uploading and emulation. Nectome is focused on just the preservation side. They have these amazing results, publish the results, they published the results. It's wonderful to see real competition in the cryonics space or the preservation space. Nectome works isn't quite the same way as 21st Century Medicine, which works on vitrification. Nectome is more focused on a type of chemical preservation. The cryopreservant is basically at or near death, you're replacing the blood supply with something that infiltrates the cells and keeps the water in the cells from crystallizing and destroying the structure in the cells. It's a chemical technique, a little less focused on vitrification. The whole point of vitrification on the 21CM side is ensuring that ice crystals don't form and that there isn't strong osmotic pressure that causes cells to explode. Nectome's more important result is demonstrating scaling this process up to an entire mammalian brain, a large mammalian brain, not even a mouse brain. We're finding ourselves in a near future/present where we have enough data to be confident that entire mammalian brains are being preserved. This immediately raises the question: why are all of the cryonics patients? Why don't we have billions of people now that we have a growing body of evidence that brain structure can be preserved by whichever technique, whether it's Nectome on one side, 21CM on the other? Sign up for cryonics. This is the way you get to see the 23rd century. I heard from the head of Alcor after my last call to action to do cryonics, lots of people flooded into Alcor. It's a nonprofit. I make no money off of saying this, no financial trust. Just get yourself a cryonics plan as part of a portfolio for longevity.
Thank you to Nick Singh and Dana Khan, our producers for supporting us on this new format. I enjoyed it. It feels organized and fun. It actually feels fun to think through the topics. It's like an entire episode worth of AMA with ourselves. It changes the stories we cover because we normally go through the most important stories to change your life. But here, when you put it into themes, you actually dig up other stories that are related to the primary topic that you otherwise would have missed.
Where should entrepreneurs actually run their AI compute? Local hardware, AWS cloud or iPhone. From Frank Gerard Marketing. There isn't a good answer. There are lots of decent answers. Benefits to each.
- Local hardware: Greater control, greater confidentiality and data privacy. On the other hand, end up maintaining said local hardware, worry about your own backups, can be a pain.
- AWS or other public clouds: Don't have to worry about that, it's abstracted away. On the other hand, compete ferociously for access to GPU resources, competing with other tenants, might have greater surface for attack depending on OPSSEC and cybersecurity familiarity.
- iPhone: Ultimate edge device until all of us are running foundation models on our watches and smart glasses, which is already happening and going to be more evenly distributed. Even greater privacy.
This shouldn't be viewed as a black or white trade-off. There's a spectrum from edge compute to data centers at the core. The best answer is to run my AI compute in the Dyson swarm, perfect blend when fully realized in a few years of data centers with 100 kilowatt nodes filling the sky, incredibly elastic. If disassembling the moon to fire off new 100 kilowatt nodes in the Dyson swarm fabric.
I spent the whole weekend dorking around on Amazon AWS Bedrock, which is a great choice. Getting started on Amazon Bedrock is a brutal process, but it does the critical thing: it captures all of your prompt history for you and any teammates into easy to manage S3 buckets. So your AI can analyze everything that you've done, which is critically important. Don't just start running on some random hardware and then lose all the prompt history.
Are humanoid robots overengineered? Would it be more efficient to isolate basic needs like food, water, and clothes and automate those directly instead? From goite FB3GN. Short answer, yes, absolutely. Why are we putting all this energy into humanoid robots? AI came into the world almost overnight and we're in a race to capital right now. What's critical for all these projects and startups is getting funded. Humanoids are so much more visually appealing that they're easier to fund. They're also easier to recruit into. That'll unlock the supply chain of all the parts and that'll unlock all of the other robots that farm and create clothes and whatever which will probably not look all that humanoid. When you look at the Gigafactory, vast majority of the automation there is not humanoid robots. It's machines that look like machines doing their jobs. Humanoids just operate those machines. They are overengineered and overinvested relative to where we'll end up, for a very good reason. Visual appeal and capital raising are a core part of this step function we're living in right now.
How can AI be used to end a war? Not as a weapon but as an impartial negotiator that all parties could trust. From jnind 5. Powerful tool. If you haven't used a large language model for negotiations, we don't know how to think other than the way we know how to think. Being able to put yourself in the mindset of another individual is extraordinarily powerful. Example: I'm anti-guns, my neighbor loves guns, can you help me explain to them my feelings in a way that lands with them and isn't viewed as offensive? You can get extreme support on your negotiation skills. Most exciting unexplored applications for AI because the system can ingest every peace treaty, every negotiation script, every conflict resolution framework that's ever been had and model outcomes with no tribal allegiance. One of the biggest challenges as humans is cognitive biases and tribal biases. Can we use this for negotiation? Absolutely. If you set as an objective function that you want to reach a balanced solution that both sides have and both sides are using AI, even different models, the probability of getting to a solution is much higher. We are biased when dealing with humans. In psychological therapy, you feel judged telling innermost thoughts to a human but not to an AI model. Real value to be had here. I'm seeing this more and more in commercial negotiation. Two parties at loggerheads, one brings in a frontier model and asks what the commercially reasonable outcome is. Bring it to the other side. The other side consults their model and they come to rapid agreement. I'm seeing this happen now over and over again.
Once AI solves most diseases how soon will treatments be available to everyone? Will access lag? From Katis 896. When do I think AI has a decent chance of solving most diseases? Timeline: optimistic elements of CZI, Chan Zuckerberg Initiative, Biohub, ARC Institute, maybe Anthropic on a good day say 5 years from now. 5 years is more rapid than historical clinical trial process end to end through three phases. If tomorrow a frontier lab says here are the cures to the top 5,000 unsolved or untreatable diseases with vast computational experiments demonstrating to the satisfaction of all experts that these are the cures or treatments, how soon broadly available? Under present regime, still multi-year process. FDA has announced two major developments:
- Adopt a Bayesian perspective as opposed to frequentist, willing to incorporate evidence from outside a particular drug for clinical approvals. Huge sea change, drug approvals can operate much more quickly taking into account pre-existing information.
- Move from two-clinical trial process to one-clinical trial process for certain cases, expediting approval.
Fast forwarding, if a frontier lab said here are the top 5,000 cures, we would see similar developments from FDA to a zero clinical trial model given enough Bayesian evidence and computational evidence. So much political pressure that we would see relatively fast availability barring exceptional circumstance.
If you had to choose one public company to bet on in the age of AI, which one and why? From Matthew Johnson 6525. Can't give investment advice, but go to 13f.info, look up the Situational Awareness Fund, which is Leopold Aschenbrenner, every quarter he files his holdings. He's killing it because he listens to what Alex is always saying: look for the innermost loop. Tailwinds in equities and assets are like nothing you've ever seen, but you have to be in the AI loop to be relevant. All Leopold's holdings are in the centerpiece of the innermost loop: chip fabs, power, chip design, algorithmic directly deriving use cases. Look at his holdings and generalize from there and you'll find lots of great stuff. W2 income is going to get pummeled in this next three years, but assets, holdings and ownership is going to go through the roof. Buy stuff, whether equities public or private, real estate, things that'll appreciate, in this next three-year window.
Outro music by CJ Truheart, piece he developed for the Abundance Summit called Moonshot Minds. Lyrics: Not because it's easy. That's the moonshot dare. Bold beat safe. We're already there. Moonshot mind. We're riding a future code. Shoot for the moon. Paving the untraveled road. The future's not a fantasy. We're making it appear. We build what was said impossible. Year after year mind. The future's what we build when you're bold enough to start. A billion lives get better when you think beyond the wall. Sci-fi turns to science facts for those who hear the call. Where the skeptics see the ceiling, we see stars to reach. Every exponential future starts with those who are dead at astral build fast and learn to fly. Smash the spot where sci-fi meets the sky. Moon shot mind. We're riding a future cold. Shoot for the moon. Baby travel road. The future's not a fantasy. We're making it up here. We built what was impossible. Year after year our minds away the future won't be built when you're bold enough to stop.
Video gets better and better all the time. AWG DB2, this was fun. Wishing you guys an extraordinary day. Salem who's airborne to Brazil. Safe travels, buddy. Soon we'll be putting you on rocket rides to get you there or hyperloop under the Gulf. Thank you to everybody for listening. Please join us if you've not subscribed, we're putting out these podcasts on a cadence that you want to get alerts. Our mission help you see an abundant world. The world is getting better at an extraordinary rate. The technologies to solve the world's biggest problems. If you're an entrepreneur, thank you for being an entrepreneur. Entrepreneurs are individuals who find juicy problems and solve problems. The more entrepreneurs on the planet, the better Earth and humanity is. Gentlemen, until I see you next time, be well. Peter D. Diamandis, your host, signing off.
If you made it to the end of this episode, I consider you a moonshot mate. Every week, my moonshot mates and I spend a lot of energy and time to really deliver you the news that matters. If you're a subscriber, thank you. If you're not a subscriber yet, please consider subscribing so you get the news as it comes out. I want to invite you to join me on my weekly newsletter called Metatrends. We spend the entire week looking at the meta trends that are impacting your family, your company, your industry, your nation. And I put this into a two-minute read every week. If you'd like to get access to the Metatrends newsletter every week, go to diamandis.com/metatrends. Thank you again for joining us today.
Keep Peter H. Diamandis in your library
Save the videos and channels worth coming back to, and find them again in one place.
