Latest / Deep Values: Value Investing Deep Dives / Naval Ravikant: The Shifting Logic of Violence and Technological Renaissance
Transcript
- 0:00Welcome to the deep dive. You know, for a long time the
- 0:03Internet had this running joke that nothing ever happens.
- 0:06Right, the whole end of history vibe, like we're just stuck in a
- 0:09loop. Yeah, exactly.
- 0:10Just, you know, waiting for the future to arrive and wondering
- 0:13why technology seemed to stall out at like, slightly better
- 0:17phone cameras. But post COVID, that era is
- 0:20definitively over. Oh.
- 0:22Absolutely. I mean, we were living in this
- 0:24period of unstable equilibrium. It looked calm, but the tension
- 0:28was building and the pandemic just shattered it.
- 0:31Yeah, it's a profound cultural and technological phase shift.
- 0:34And so our mission today is to explore that shift.
- 0:37We're pulling insights from a recent conversation featuring
- 0:40Naval and we're looking at how AI is fundamentally
- 0:43restructuring everything, like from the death of the corporate
- 0:46ladder to the sudden restructuring of global warfare.
- 0:49It's heavy stuff that it's all compounding so rapidly right
- 0:52now. It really is, but before we get
- 0:54into it, I just have a quick message for you to access more
- 0:57advanced researches, applications and our latest
- 0:59published personal intelligent multi agent deep value
- 1:02intelligence. Welcome to subscribe on Deep
- 1:04Values dot AI. And if you want to track total
- 1:07market temperature, welcome to search and download the iOS app
- 1:11Buffett Indicator from the Apple App Store.
- 1:13Right. And just a quick disclaimer
- 1:15here, all analysis is based on publicly available sources and
- 1:19maybe subject to revisions or differing interpretations.
- 1:22All contents from the channel is for educational and research
- 1:25purposes only and does not constitute investment advice.
- 1:29Yeah, definitely conduct your own due diligence before making
- 1:32investment decisions. For sure, and if you like to
- 1:34access wider topics including life and investing philosophy,
- 1:37please subscribe to our sister channel at Deep Values.
- 1:40Don't forget to like this video. Much appreciated.
- 1:44OK, let's unpack this. We're seeing sci-fi
- 1:46technologies. The drones, the AI rockets.
- 1:49Yeah, they're no longer fiction. They are incredibly high demand
- 1:53and we're kind of lacking the what Naval calls the sci-fi
- 1:57engineers to build them. Well, and that changes how the
- 1:59basic unit of work is organized. Right, so let's talk about the
- 2:02corporate ladder, or I guess the end of it.
- 2:05The source material talks about Naval Zone Company.
- 2:08Impossible. And get this, there is no Slack
- 2:110, no complex project management software either.
- 2:15It's literally just GitHub and direct text messaging.
- 2:18Wow. Yeah, which sounds insane if
- 2:20you're used to the traditional corporate hierarchy.
- 2:23Like the old system is essentially a tree, right?
- 2:25Exactly. You have the CEO at the trunk,
- 2:27middle managers on the branches, and workers at the leaves.
- 2:31And it's basically a rigid, high stakes game of telephone.
- 2:34Yes, which is slow. Naval prefers what network
- 2:37theorists call a fully interconnected graph.
- 2:40Right, but doesn't that just lead to total chaos?
- 2:42I mean if everyone is just texting everyone else without a
- 2:45manager. Well, that's the thing.
- 2:46A fully interconnected graph only works if every single node,
- 2:50meaning every employee, is highly intelligent and
- 2:54autonomous. So it's an aggressive filter.
- 2:56Completely. If they can't navigate the
- 2:57chaos, if they need someone to mediate a dispute or assign them
- 3:01tasks, they just don't belong there.
- 3:03They need the tree structure. But even with a team of
- 3:05geniuses, a massive software code base is still a maze.
- 3:10Like how does a coherent product actually ship without someone
- 3:13managing the dependencies? And that's where the phase shift
- 3:16really happens. AI is the implicit glue.
- 3:20It's the implicit middle manager.
- 3:21Oh, wow. Yeah, they call it vibe coding.
- 3:24Instead of typing syntax, an engineer uses natural language
- 3:27to just tell the AI to write, test and integrate the code.
- 3:30So if I'm an engineer in Tokyo and I write something dense on
- 3:34Tuesday. And then someone in New York
- 3:36logs on and needs dild on it. Right in the old system, New
- 3:39York has to wait for Tokyo to wake up or ping a manager.
- 3:44But now you just unleash an AI agent on the repository, you say
- 3:49parse that commit history and explain it to me.
- 3:51So it acts as like an HR directory too.
- 3:54Yeah, exactly. It finds the experts within the
- 3:56Curbase. What's fascinating here is that
- 3:58AI allows for on demand reporting.
- 4:00Right, instead of building those static backward looking
- 4:03dashboards. Exactly, you just prompt the AI
- 4:05to act as a real time data analyst.
- 4:07You can generate a Got chart or track project delays instantly,
- 4:11completely eliminating traditional management overhead.
- 4:13Which is wild. So if AI is doing all this glue
- 4:17work for software, making it frictionless, it kind of
- 4:20fundamentally changes the relationship between software
- 4:24and the physical world, right? No.
- 4:25Absolutely. It ushers in a massive
- 4:27renaissance for hardware. Right, because historically,
- 4:30building the physical device was only half the battle.
- 4:33Usually less than half. You could build a great piece of
- 4:36hardware, but it was basically a brick without custom APIs or a
- 4:40dedicated software team to build the interface.
- 4:43And the hardware and software teams spoke completely different
- 4:46languages, but now the lines are blurring.
- 4:49Because the AI allows a hardware engineer to do like 20 to 30% of
- 4:54a software engineer's job. Which is just enough to bridge
- 4:56the gap. Exactly.
- 4:57You don't have to wait months for a custom API.
- 5:00The hardware creator just uses AI to make the device functional
- 5:03immediately. And this leads into a really
- 5:05surprising geopolitical dynamic that the source material brought
- 5:08the push for open source AI. Yeah, the open source Trojan
- 5:11horse. Right, because you have these
- 5:14massive entities like NVIDIA and even entire manufacturing
- 5:17powerhouses like China actively pushing for AI software to be
- 5:22open source and free, which sounds counterintuitive.
- 5:25It does, until you look at the economics.
- 5:28It's the classic strategy of commoditizing your compliment.
- 5:31Explain that a bit more. Well, if your primary profit
- 5:34comes from hardware, you want the software that runs on it to
- 5:37be free. NVIDIA and China actively want
- 5:40AI to be free. Why?
- 5:42Because commoditizing software makes hardware the ultimate
- 5:46value driver. Oh, I get it.
- 5:47It's like if you sell hot dogs, you want the buns to be dirt
- 5:50cheap. Exactly.
- 5:51Free buns mean you sell way more hot dogs.
- 5:54If AI software is free, you sell more chips and you sell more
- 5:56physical consumer goods. So what does this all mean for
- 5:59you, the listener? Well, if you want to build a
- 6:01smart security camera or I don't know, programmable lamp.
- 6:04The software barrier to entry just dropped to basically 0.
- 6:07Right. You don't need venture capital
- 6:09able to hire a software team anymore, so building things is
- 6:12easier than ever. But that transitions us into
- 6:14some big AI unknowns, like who ultimately controls the
- 6:18underlying intelligence. And that's where the
- 6:21democratization friend hits a wall.
- 6:22Yeah, because we're going from decentralized consumer hardware
- 6:27to massively centralized tech monopolies training these AI's.
- 6:32Right. The current conventional wisdom
- 6:34is that AI training is becoming highly centralized, dominated by
- 6:38maybe 2 to 4 mega companies. Because they're the only ones
- 6:41who can afford the gigawatts of power and the data centers.
- 6:44But I have to ask, what if the conventional wisdom is wrong?
- 6:49Like, is distributed training the contrarian bet we should be
- 6:52looking at? It's a nice thought, like using
- 6:54millions of laptops hooked together to train a model, but
- 6:57physics is kind of in the way. Because of latency, right?
- 7:00Exactly. When you're updating weights in
- 7:02a frontier model, the processors need to talk to each other in
- 7:05microseconds. The latency over the public
- 7:07Internet between a laptop in London and a server in Texas
- 7:10just breaks the training run. So you really do need them all
- 7:13physically sitting next to each other, connected by fiber
- 7:16optics. Yeah, but while the training is
- 7:18centralized, we need to bring a little reality to the AGI hype,
- 7:22the artificial general intelligence hype.
- 7:24Oh for sure. The source points out that
- 7:26current AI has what naval calls jagged intelligence.
- 7:30Jagged intelligence. That's a great term for.
- 7:32It it is because it can, you know, write a graduate level
- 7:36essay on history, but it struggles with basic multimodal
- 7:39reasoning or simple logic that a kid could do.
- 7:41This raises an important question about world models,
- 7:44which gets so much hype online. Yeah, you see those AI videos
- 7:48where you're flying through a photorealistic 3D room and
- 7:51everyone online goes, Oh my gosh, the AI understands
- 7:54physics? But it doesn't.
- 7:56Generating a sequence of pixels that looks like a 3D World you
- 7:59can walk through does not mean the AI has an internal world
- 8:02model. Right, it's just predicting the
- 8:04next pixel based on patterns. It doesn't have an internal
- 8:07reinforcement learning loop to actually predict consequences
- 8:10and take actions. Exactly true world models where
- 8:13the AI actually understands why a glass shatters when it hits
- 8:16the floor. Those are still emerging.
- 8:18OK, so we've got these incredibly powerful yet somewhat
- 8:21jagged tools. And this brings us to a much
- 8:24darker transition, because if individuals are empowered to
- 8:28build software and hardware easily.
- 8:30They're also empowered to create destruction.
- 8:33Right. Democratized destruction.
- 8:35The historical parallel they drew in the source was
- 8:37absolutely fascinating. The history of violence.
- 8:40So if you go back, rifles democratized power.
- 8:44Right. Yeah, a peasant with a rifle
- 8:46could shoot a knight in full armor.
- 8:48Exactly. And because of that, nation
- 8:51states had to rise up to create centralized factories and
- 8:53training to manage those rifles. Then you Fast forward to nuclear
- 8:57weapons and power centralizes again to just 7 to 9 sovereign
- 9:01nations. Mutually assured destruction.
- 9:03Right. But here's where it gets really
- 9:05interesting. Drones.
- 9:07Drones bring mutually assured destruction down to the
- 9:10individual level. It's terrifying, honestly.
- 9:13The tactical advantage is completely inverted.
- 9:15Because the attacker has all the advantages.
- 9:17Kinetic energy, gravity, mass. Surprise.
- 9:20You can launch a swarm of cheap off the shelf drones powered by
- 9:24that free open source AI we talked about earlier.
- 9:26And the defender? Their only advantage is short
- 9:28range, having to shoot up instead of flying all the way to
- 9:30the target. But economically, the defender
- 9:33is ruined. Right, shooting down a $500
- 9:35drone with a $2,000,000 missile is a losing and.
- 9:38This logic of democratized violence isn't just hardware.
- 9:41It bleeds into biology, which the source material covers
- 9:45pretty carefully. Yeah, just as vibe coding let's
- 9:47anyone write software, AI might soon let thousands of people
- 9:51engineer biological threats. The source mentions the Wuhan
- 9:55lab leak theory impartially of course, just as an example of
- 9:58how a localized biological event can halt the whole world.
- 10:01But historically, engineering a pathogen took a state level lab.
- 10:05Now AI reads the genomic data for you.
- 10:08But wait, AI could also cure these threats right?
- 10:10I mean, it can design the vaccine just as fast.
- 10:12Theoretically, yes, but we face a massive regulatory bottleneck.
- 10:17The source argues pretty strongly that bioethicists and
- 10:20heavy regulations prevent rapid testing.
- 10:23So instead of letting voluntary informed people test cures
- 10:26quickly. The bureaucracy traps the data
- 10:28we need to save lives in silos. Bad actors don't have to wait
- 10:33for FDA approval to release a threat, but the good guys have
- 10:36to wait 10 years for clinical trials to release a cure.
- 10:39Men, that is, that is incredibly bleak.
- 10:42Drone warfare, Centralized tech monopolies, Democratized
- 10:47bioweapons. How are we supposed to process
- 10:49all this without just totally panicking?
- 10:52Well, this is where we have to talk about the case for
- 10:53irrational optimism. OK, please I need some optimism
- 10:57right now. The problem is a failure of
- 10:59imagination on our part. We easily fixate on doom because
- 11:03it's highly legible to our brains.
- 11:05Legible, meaning it makes logical sense.
- 11:08Right. It's very easy to predict how
- 11:10jobs will be destroyed by AI, but predicting the entirely new
- 11:14industries AI will create? That requires immense creativity
- 11:17that we usually lack. Yeah, it's like 200 years ago,
- 11:19almost everyone worked on a farm.
- 11:21Exactly. If you told an 1800s farmer that
- 11:24machines were going to take 90% of the farming jobs, they would
- 11:27assume mass starvation in the end of the world.
- 11:29We couldn't have imagined that their descendants would be
- 11:32software engineers or podcasters.
- 11:34Right. No one could have predicted 90%
- 11:36of the jobs we have today. The doom scenarios from a
- 11:38century ago sound exactly like our doom scenarios today.
- 11:42If we connect this to the bigger picture, the core philosophy
- 11:45from Naval is that we must be irrationally optimistic.
- 11:48Because pessimists don't build anything.
- 11:50Right. He compares pessimists to crabs
- 11:52in a bucket just dragging people down.
- 11:54They offer no actual solutions. Optimistic is basically the only
- 11:59actual way forward through these, well, these interesting
- 12:02times. It's an active choice.
- 12:04So to sort of summarize the journey we just went on, we
- 12:08looked at the death of the corporate ladder, right,
- 12:11replaced by interconnected graphs and AI middle managers.
- 12:14You saw the Renaissance of hardware and open source AI.
- 12:16We looked at the grim restructuring of global violence
- 12:19with drones and biology, and we landed on the absolute necessity
- 12:23of optimism. It's a lot to take in, it is.
- 12:26But I want to leave you with a final, slightly provocative
- 12:29thought to Mull over something building on all this.
- 12:32If AI inevitably commoditizes coding, managing, and even
- 12:36complex problem solving, what if the most valuable human skill of
- 12:40the 21st century isn't technical at all?
- 12:43Oh, that's a great point, right? What if the ultimate currency of
- 12:47the future is simply the creative optimism required to
- 12:50imagine a positive use for these God like tools?
- 12:53That really might be the only bottleneck left our own
- 12:56imagination. Exactly the era of nothing ever
- 12:58happens is over. So thank you so much for joining
- 13:01us on this deep dive. Take these ideas into your week
- 13:04and, you know, try to stay rationally optimistic.