EP32:The Thinking Machine: Inside Jensen Huang's AI Empire
33:07 Share

Welcome you. So you've sent over this really rich set of materials on Stephen Witt's book, The Thinking Machine. Yeah, it's a great selection. Chapters 8 through 34 really gets into the meat of the story. Right. And our mission today is basically to pull out the most important stuff, the really fascinating insights from these chapters. Give you a clear picture of the key themes, the incredible journey Witt lays out. Exactly. Because

Because there's a lot here from, you know, philosophy on pushing limits to the super intense world of AI development. Absolutely. We're going to connect some surprising dots, I think, from Marcus Aurelius to Jensen Huang. Okay, let's jump in then. Chapter eight opens with that epigraph. Practice even what seems impossible.

Marcus Aurelius. Gets you right away, doesn't it? It really does. Makes you immediately think about like mental barriers. Yeah. Self-imposed limitations. Yeah. And Witt connects it directly to tackling these huge daunting projects. You know, the feeling of being overwhelmed. Mm-hmm.

There's that anecdote in the book about breaking down a massive task, right? Right. Into smaller pieces. So you don't get paralyzed by the scale of it. And I love the analogy he uses, training a muscle. You got to work at the hard stuff, the impossible stuff, to build that mental strength. So the takeaway is pretty clear. Lean into the difficulty. Practice it. Don't shy away. Okay, so chapter nine introduces Jensen Huang.

And he's presented as much more than just, you know, the CEO of NVIDIA. Oh, definitely. It's framed as this epic journey, right? Started with video game hardware. And ends up building the world's most valuable company. It's wild. That 30-year vision is incredible. And he comes across as, well, unconventional. A thinker. A gambler.

A bit of a renegade, maybe. And CEO for 32 years longer than anyone else in the S&P 500 currently. That staying power. And the personal details Witt includes, like meeting him at Denny's,

his favorite place yeah that detail really grounds him makes him seem more relatable despite the huge success the jokes the story about being a dishwasher right dishwasher and bus boy it really paints a picture of his journey born in taiwan u.s at 10. denny's was his first job it gives him that sort of immigrant outsider perspective which seems important and nvidia's start in 93 purely high-end gaming graphics at first but he didn't stay there he made that

a huge bet on parallel computing. Which was super risky. Others had tried and failed, right? Exactly. But he saw something. And parallel computing, just to clarify, is like having many processors work on a problem simultaneously instead of just one working really fast. Gotcha. And he found customers outside gaming pretty early on. Yeah. Weather forecasting, radiology, oil exploration, showing that tech had broader uses. Then comes 2012, that pivotal moment.

NVIDIA's gaming cards are suddenly useful for neural networks. Which, as Witt points out, were kind of a joke in the AI world back then. Fringe stuff. But Huang doubled down. He saw the potential. Incredible foresight. Seriously, and look where they are now. NVIDIA's chips are basically the engine behind ChatGPT, all the big AI models. Demand is insane. And Witt stresses, he's still an engineer at heart.

driven by not failing. And he's optimistic about AI, sees it as a force for good, this new industrial revolution. He uses that microwave analogy. Right, trying to calm the fears. Like, it was scary at first, now it's just a kitchen appliance. But what really drives him, according to the book,

Fear of failure. Still, even after decades. That Nate to prove himself. Working 12, 14 hours a day. It's intense. So his story is really about like amplifying core principles, hard work, courage, mastery to achieve something extraordinary. Okay, chapter 10, part one, just the title, The Thinking Machine. Yeah, very stark, very simple, but it immediately makes you pause. Makes you ask, what is a thinking machine? What is thought?

Intelligence. Human. Artificial. And Witt notes the deliberate layout, just the title. It's meant to provoke, right? Sets a philosophical tone, maybe scientific, maybe even speculative. It's designed to spark curiosity, not give easy answers up front. A deliberate setup. Which leads nicely into chapter 11, which is chapter one of the book, titled The Bridge, The Thinking Machine. And here we get Jensen's early life in detail. Yeah, that arrival in rural Kentucky. Age 10, 1973. Limited English. Total culture shock.

from Taiwan and Thailand to tobacco farmers and coal miners.

a different world. And that image of the footbridge, terrifying, missing planks, bullies waiting. But his resilience shines through, laughing it off almost. His classmates saying he seemed to be having fun, wild. It really speaks volumes. And becoming the top student despite the language barrier, teaching his friend to fight back shows that grit early on. We learn about his parents, teaching them English from the dictionary, the tough decision to send him and his brother away due to political unrest. And

And ending up at Oneida Baptist Institute, which sounded less like a boarding school and more like, well, a reform school. That story about his roommate getting stabbed and Jensen learning to bench press from him. Just incredible adaptability in harsh circumstances. Then high school in Oregon. Nerdy clubs, Apple II obsession, but also this amazing table tennis talent. National ranking, super fast. Again, that intense focus, that work ethic applied to mastering something totally new.

And back to Denny's working there, methodically eating through the menu. His belief that he thinks best under adversity. Plus the academics skipping a grade. Great GPA.

Choosing Oregon State with his friend. Meeting Lori. Right, in electrical engineering. Winning her over with homework helped his superpower. It all leads to graduation, the Silicon Revolution hitting its stride, and his move to Silicon Valley. Witt ties it all together how these early, sometimes tough experiences forged his resilience, adaptability, focus, and curiosity. Okay, chapter 12, chapter 2.

Large-scale integration, the thinking machine. His early Silicon Valley days starts very dramatically. Yeah, that scene driving in the snow after proposing to Lori, AMD Christmas party in 1984. He's got the job. He's confident. And then, bam, black guy's accident. Months in a neck brace. But Witt notes it strengthened their relationship. And his work at AMD sketching microchip designs by hand, it really highlights how manual it was back then. Almost like art.

Learning Mandarin from the photo mass workers to always learning, connecting. Then the move to LSI logic. This is where large scale integration, LSI and then VLSI, very large scale integration really took off. So cramming way more components onto chips, automating the low level design stuff, a huge industry shift. And his work ethic there was insane. No weekends. But also, crucially, that ability to pivot fast when things weren't working. His colleague Jens Horstmann mentions this.

That balance is interesting, intense focus, but also risk-taking. Signing deals for unproven tech and then making it happen. While also building a stable home life with Lori, marriage, house, kids, amidst the chaos. Then the key meeting, Chris Malachowski and Curtis Prime from Sun Microsystems. They see his talent. They have this idea, a cheaper graphics chip for PC gaming. Sun says no.

So they decide to go it alone and they want Jensen to lead. But he doesn't just jump. He researches, talks to customers, analyzes reports, methodical. Witt ends it feeling like Jensen's rise was almost inevitable. Talent, timing, leadership, that foresight. Okay. Chapter 13, chapter three, a new venture, the thinking machine, the founding of NVIDIA, not in a garage. Nope. At Denny's in San Jose. Next to a freeway, lots of cops. Witt calls it a mythic origin story. Definitely contrasts with the usual startup tale. And the context is key.

Big players ignored PC gaming. Huang saw the gap. Even though there were, what, 35 competitors already? John Petty warned him off. But Huang kept crunching the numbers on his spreadsheet. Persistent. And the name story is hilarious. Envision was taken by a toilet paper company. Yeah, so they landed on NVIDIA from the Latin for envy. Classic. Then that bizarre scene finalizing things at Denny's with bullet holes in the window. Ah!

And rushing to Preen's condo to sign. Adds a bit of gritty reality to the founding myth. Then the actual work starts in Preen's condo garage, Huang's office, a small circular table. And the first chip, the NV1, packed with features, but incompatible features, right? Quadratic texture mapping. Yeah, totally out of sync with what game developers were using, which was mostly triangle-based.

A fundamental mismatch. So the launch was... A disaster. Glitches, clipping, crashes, everything went wrong. Despite Wong's grand architectural vision drawn on the walls. Right. Huge setback. They'd built a whole supply chain around this flawed ship. But Witt concludes, this failure taught them the crucial lesson. Build what people actually want. Set them up for later. Chapter 14-4. 30 Days the Thinking Machine.

Witt calls this a masterclass in survival. David Kirk's first impression in 96 paints a grim picture. Ghost town office, demoralized staff, feuding founders, that weird hardware emulator. Yeah, things look bad. NVIDIA was almost broke after the NB1 failure. Jensen's plan, drastic. Pivot to Microsoft, ditch Sega, rush out a knockoff chip. And slashing the team, over 100 down to 35 engineers. Imagine the atmosphere. Then the craziest part.

Skip prototyping, go straight to mass production using the emulator. Unheard of. That emulator sounds like it was held together with tape and prayers. Slow, clunky. But it worked. Just barely. Engineers like Bill Diericks and David Kirk pulling double shifts, testing the new NV3 design in agonizing slow motion. While tensions rose between Jensen and Curtis Prime technical vision versus sheer survival. And Jensen's management style.

Those public learning opportunities sounded pretty harsh, almost verbal abuse. That story about the guy asked to refund his salary for a mistake. Oof.

Tough environment. People left if they couldn't handle it. But they got the NV3 design done. Weeks of grinding. Celebratory beers. But Jensen was still unsure. 50-50 chance, he thought. It worked. Flawlessly. The Riva 128, huge hit. Sold a million in four months. Saved the company. Wow. And this experience changed Jensen. Led to that risk-taking, that mantra. Our

Our company is 30 days from going out of business. Always on the edge. The takeaway. Sometimes you get to break the rules, take that leap, especially in a crisis.

Okay, Chapter 15-5, going parallel, internal conflict time. Yep, that intense showdown in 1998, Jensen versus Curtis Prime, leads to Prime stepping back. And David Kirk becomes Jensen's key tech guy. Plus, they started aggressively poaching engineers from rivals, building the A-team. And Huang was obsessed with Christensen's innovator's dilemma, focusing on that new PC gaming market the big guys ignored. Smart.

Then the Reva TNT chip, the breakthrough with twin pipelines for parallel processing. Right. Huang had been skeptical about parallel, but realized games were getting too complex for the old way. And getting John Carmack, the legendary programmer, to optimize Quake 3rd for it, huge validation. Massive. Wit ends hinting this parallel stuff had implications way beyond gaming. A bigger shift was coming. Chapter 16-6, Jellyfish.

fascinating sidebar about AI and backgammon. Yeah, contrasts the famous Deep Blue chess match with this lesser known but maybe more significant backgammon program, Jellyfish in '97. Because Deep Blue was brute force, right? Just calculation power. Exactly. Jellyfish was different. It was a neural net. It learned by playing itself. It came up with new strategies humans hadn't thought of. Enter Frederick Dahl, the creator.

a Norwegian Renaissance man of nerdy hobbies. Yeah. Inspired by making computers think like brains. And the name? Jellyfish. Nodding to simple nerve nets. Even just a hundred artificial neurons were enough to beat top humans. And it wasn't just a game. It changed how humans played backgammon. Experts studied it, learned from it. Right.

But Dahl tried applying it to poker. No dice. Too complex. The bluffing element. Showed the limits back then. Witt uses it to show where the real AI breakthroughs were happening. Learning, adaptation, innovation, not just raw calculation power. Okay, back to NVIDIA. Chapter 17, 7. Deathmatch.

NVIDIA's rise mirrors the rise of pro gaming. Enter Jonathan "Fatal" Lundy Wendell, dominating quick thrower around 99, intense practice, almost athletic. And he used NVIDIA's TNT2 card, gave him a crucial edge in frame rates. Witt compares it to Jordan and Nike.

performance gear. Meanwhile, Nvidia was exploding new cards every six months. IPO in 99, stock goes nuts. But Jensen stayed focused on crushing the competition, especially 3DFX, sending demanding emails right after the IPO. Ruthless. Nvidia just outplayed them. Better engineering, better marketing. Coining GPU for their GeForce cards was brilliant branding. Eventually, 3DFX folds. Nvidia buys their best engineers for a million bucks each, though the engineers weren't exactly thrilled, called Juan Darth Vader. Shows the culture clash.

NVIDIA was all about execution, speed, even if the code wasn't elegant. The 3D Feifex guy called it cancer. By 2001, NVIDIA hits $1 billion in GPU sales, powering huge games like Halo. Wit touches on Huang's paranoia, his drive, no family hires, meticulous about details like moving his pool house. And that risky move adding programmable shaders to GPUs paid off massively. More flexibility for developers.

The chapter ends with NVIDIA joining the S&P 500, replacing Enron, funnily enough. Huang's a paper billionaire. They're established, but volatile times ahead. Chapter 18. Eight. The compulsion loop. Navigating the dot-com bust early 2000s. Rough times. Oh, yeah.

stock tanked the noisy geforce fan debacle the xbox deal failed and sec investigation problems everywhere wit highlights huang's contradictory nature here intense ceo explosive temper that geforce fx screaming incident which somehow motivated people weird but then also the warm down to earth side reconnecting with his college buddy focusing on family and nvidia strategy cater to those obsessive pc gamers understanding that compulsion quest rewards upgrades

Gotta have the latest card. Their marketing with the CGI pixie Don encouraging custom rig building. It built a loyal base. And despite the crises, they kept growing. Record earnings by 2004. The gaming market saved them. But Wall Street was still unsure, right? Because Juan kept reinvesting in speculative tech. Exactly. The chapter ends teasing his next even crazier bet. Setting up the next big thing. Which brings us to Chapter 19-9.

CUDA, the accidental revolution, starts with Ian Buck's crazy experiment. Trying to max out Quake 3 visuals by linking 32 GeForce cards, 8K resolution, insane. And that's when he realized these GPUs had way more processing power than games needed. Potential for science. Stuff that would take humans millennia to calculate by hand. But there was a problem. NVIDIA's code only understood triangles.

for graphics. Right, so Buck dives into shading textbooks. Ironically, gets bored with graphics, fascinated by the raw hardware power. Leading him to create Brook? Open source language. Let scientists use GPUs for simulations, galaxies, nuclear bombs. And Jensen sees this, brings Buck into NVIDIA. Develop it further, that becomes CUA. So the core idea.

Repurposed GPU parallel power for science. Turned gaming cards into supercomputers. Enter John Nichols, intense engineer behind CDA. He saw Moore's Law slowing down for traditional CPUs. Nichols saw CUNY as the way past Moore's Law.

through parallel processing, thousands of cores working together. But initially, no demand. Witt mentions the first customers are just two breast cancer researchers, a $0 billion market. Yet Wang believed, saw the future with massive data sets. Big data was coming. CEA would be essential. Still, adoption was slow after the 2006 launch, hard for programmers. Then BumpGate hit. The failing laptop GPUs cost him $200 million in refunds.

Ouch. Yeah, huge setback. But Huang kept investing in CUD. Didn't waver. Wit-framed CDA not just as tech, but as a vision they stuck with. Persevered through the tough times. Laid the groundwork for everything AI. After 2010, Resonance. Enter Bill Dally, huge figure. Incredibly accomplished guy. High school dropout to MIT professor. Patents galore. Adventurous hobbies. Renaissance man. And he joins NVIDIA, which was still seen as kind of shaky over at Intel. Surprising. But he admired Jensen's leadership. That was key.

Dally transforms NVIDIA research, broadens the focus, robotics, climate modeling, biochemistry, all powered by faster GPUs. And wit reminds us the CUDA bit was still seen as crazy. Tiny market of academics. Investors were skeptical. Activist investor Jeff Smith challenged Huang directly. But Huang held firm.

risked his job convincing them CUA was the future. Again, linking back to the innovator's dilemma, disruptive tech starts niche. Love the anecdote about Huang visiting the physicist in Taiwan using GPUs and desk fans to build a supercomputer. Shows the passion of early adopters. Witt describes Huang's approach as

Resonance. Not a rigid plan, but sensing needs, sensing breakthroughs, connecting with people. This period marks the big shift. NVIDIA becoming a leader in AI and supercomputing, built on risk and belief. Chapter 21. 11. Alex Ned. The deep learning revolution truly ignites. Meet Alex Krzyzewski. Brilliant programmer, but private.

Hinton called him the best he ever met, but neural nets were still fringe in the late 2000s AI world. Hinton had to call it machine learning or deep learning just to get funding. Alex joins Hinton's lab, partners with Ilya Sutskiver, different personalities, same huge goal, teach a computer to see using GPUs. Which nobody else was really doing. GPUs were for games. They pooled their own money for GeForce cards, set them up in Alex's bedroom. Then Alex works his magic.

builds AlexNet's 650,000 neurons, trains it on Fei-Fei Li's huge ImageNet dataset. The results were stunning. Accuracy jumped from basically 0 to 80%, crushed previous methods. Alex made it run incredibly fast on GPUs. Hinton called him a wizard. Then the 2012 ImageNet competition. Witt says a bomb went off. The results shocked everyone. Fei-Fei Li double-checked for errors. And suddenly, everyone was doing neural nets.

Deep learning took over by 2014. Alex's paper became legendary. Then Google swoops in, buys their company, DNN Research, for $44 million. Alex, Hinton, Ilya. A big bang moment for AI. Witt emphasizes timing and vision. Alex saw GPU potential for AI, had the skills, had the data. Perfect storm. Okay, Chapter 22, Part 2 again. The thinking machine, a marker, a shift. Right. Moving from the backstory into the core ideas. Reiterating the central theme.

The simple structure signals a deeper dive is coming. Unpacking the central framework. Feels like a turning point. The real performance begins, as Witt puts it. Still exploring what that thinking machine really means. After 2312, OIA-LO introduces Brian Catanzaro, a unique figure at NVIDIA. Totally. Long hair, loud shirts, calm voice, humanities degree in Russian Lit, Mormon missionary in Siberia.

Then computer engineering. Not your typical chip guy. And his Intel internship inside back in 2001, realizing Moore's law was hitting limits and Intel was maybe in denial. Very perceptive. Moves to Berkeley, co-founds the Parallel Computing Lab, saw AI as the next big thing, even when professors were skeptical, juggled internships to get

by. Lands at NVIDIA, works with the open-minded Bill Dally, who encouraged research. Brian becomes NVIDIA's first dedicated AI researcher. And his results were amazing. Replicated Andrew Nang's famous cat recognition project and used 2000 CPUs. Brian used just 12 GPUs around 2012. Massive difference. Then his work, building CUDDNN that library to speed up neural nets.

facing huge challenges: new baby, med side effects, lack of software experience. So he goes straight to Jensen. Hwang gets it instantly, writes O-I-A-L-O on the whiteboard, once in a lifetime opportunity. Full backing. CutTNN gets built fast. By 2014, NVIDIA's all in on AI. Jensen declares, "We are an AI company." Google deploys massive GPU clusters, demand skyrockets.

Witt concludes the AI revolution was this convergence, parallel hardware, GPUs, and neural net software with Brian, the humanities guy, right at the center. Chapter 2413, superintelligence. Wang's focus on AI becomes all-consuming. Yeah, Witt describes him abandoning hobbies, cooking, cars, Maui trips, less communication, totally obsessed. Driven by that conviction, AI is the once-in-a-lifetime chance, working even harder, if that's possible. And the market responded.

Stock surges. Wall Street finally sees them as an AI company. Cloud providers, Google, Amazon, Microsoft buying Jesus is like crazy for their data centers. Witt contrasts Huang's optimism with Elon Musk's warnings about AI risks. Interesting difference given their similar drive. And that symbolic moment.

Huang personally delivering the first DGX-1 AI supercomputer to Musk at OpenAI. Yeah. The chapter underscores Huang's intense focus and conviction as the engine driving NVIDIA's and the whole tech landscape's transformation. Chapter 2514. The good year. 2017. NVIDIA's breakout. Absolutely. Revenue doubled. Profits tripled. Products flying out the door. Wit touches on the rivalry with AMD.

That palpable tension between CEOs Huang and Lisa Su, distant cousins apparently. Contrasting styles. AMD gained some ground, especially from Intel. But Huang snagged a huge win, getting Nintendo to use Nvidia's Tegra chip in the Switch. Massive success. Though Huang barely mentioned it. They were moving so fast on AI. By 2017, they were the AI company, powering facial recognition recommendations, even Nobel Prize-winning science. The crypto craze added fuel to the fire, driving GPU demand.

Though Huang seemed ambivalent, prioritized gamers and scientists. And their new HQ, Endeavor. Huang deeply involved in the futuristic design. Huge scale. Thinking big. Plus launching Isaac for robotic simulation, Omniverse for digital twins, pushing boundaries constantly, and they hit a $100 billion valuation by end of 2017. Huang solidified as a visionary. But bigger things were still coming. Chapter 26, 15.

The transformer, the AI concept that changed everything. Meet Jacob Ooskerite. His 2014 idea, simplify language models, focus on context, minimize complex structures inspired by the brain. The core, context alone, replacing complex memory with a simple knowledge graph and self-attention, letting the model weigh word importance. Initial skepticism at Google, even from his dad, but it worked well on GPUs. Collaboration with Ilya Polosukhin, his opponent.

his epiphany. Self-attention could link words across whole documents, not just sentences. Big leap in understanding context. Their 2017 English to German translator improved incredibly fast with more data, seemed to learn on its own. They experimented with music, art, showed its broad potential. But Google kind of missed the boat on its value. The original team mostly left for startups by 2023. But Ilya Sutskiver at OpenAI, he saw it,

shifted OpenAI's entire focus to transformers. Led to GPT, generative pre-trained transformer, early versions, learning from vampire romance novels, funny detail. Then GPT-2, human-like text answering novel questions. Huge leap. And Sutzkever starts worrying about AGI, the alignment problem, making sure AI stays beneficial. OpenAI keeps scaling up.

GPT-1 to GPT-2 required massive compute power, led to the capped profit structure Microsoft's $1 billion investment, ends with the idea of an AI factory needed for training, sense of being at the start of something huge. Chapter 27, 16, Hyperscale, Huang's obsession with scale, NVIDIA as an AI powerhouse. His view, AI is software running on NVIDIA hardware. At GTC 2018, he announces a 25x GPU speedup in five years.

Supercharged Moore's Law. The comparison is stark. AlexNet took a week to train in 2012, 18 minutes on a DGX2 system later. Unbelievable acceleration. Kuang saw data centers as one giant chip, but the network became the bottleneck. GPUs were too fast for existing connections. Solution. Acquire Mellanox. Get their ultra-fast InfiniBand networking tech. Bottleneck solved.

Witt also mentions the failed ARM acquisition attempt in 2020. And crucially, NVIDIA's dominance isn't just hardware, it's software. CDA, math tricks. Witt says software accounts for most of that thousand-fold AI speedup between 2012-2022. Their strategy of giving away free software tools locks people into their ecosystem. Smart business. That Caltech anecdote, waiting 18 months for H100 chips because they relied on CUDA, perfectly illustrates it. Witt describes the intense culture under Dwight Diericks.

Be first, even if software isn't perfect. Ship fast, iterate. Gave them a lead. And NVIDIA handled remote work smoothly during COVID. HONG read weekly priority lists from everyone. Super responsive email, stayed connected. HONG tasks boss arts with creating a toolkit for transformers. Internal demos were showing incredible stuff. Real-time face replacement, Picasso GAN images. WIT concludes NVIDIA is always innovating, keeping things quiet. Their current capabilities are probably underestimated. Chapter 2817, Money.

the economic explosion. OpenAI's GPT models reshaping the world. GPT-3's release in 2020, massive dataset, some maybe pirated, impressive skills, logic, code, poetry.

But ChatGPT's launch in late 2022 was the public explosion. Instant popularity, endless applications, then GPT-4, even smarter. Acing exams, understanding images, designing websites from sketches. But the costs are immense. Over $100 million just to train GPT-4. Yeah. Plus inference costs for every query. Yep.

OpenAI subscription model took off though. Huge demand. And the ripple effects everywhere. Microsoft Co-pilot, DeepMind's AlphaFold 2, daily mid-journey, stable diffusion for images, Sora for video. Plus industrial AI transforming sectors like power grids, logistics, and AI even accelerating AI development itself. Like DeepMind optimizing math, NVIDIA using AI to design chips. It's recursive. But Witt acknowledges the dark side.

AI voice scams, deep fakes, needing more AI to fight the bad AI. Ironic. And NVIDIA is right at the center. Huang becomes this hugely powerful figure. Chips are like gold. Stock skyrockets. One of the world's most valuable companies. Immense wealth created inside NVIDIA. Strong sense of purpose among employees building this revolution. Chapter 29. 18. Spaceships.

Inside NVIDIA's futuristic HQ late 2023 Voyager building. WIT makes it sound like sci-fi. Open air design, central mountain, vertical gardens, even AI tracking employees for cleaning? A bit creepy. Maybe a little. But the atmosphere sounds enthusiastic.

Employees admire Jensen. Custom emojis celebrating market cap milestones. Contrasted with the tough conditions for QC engineers in noisy windowless labs testing chips to failure. WITC's amazing demos. Hyper-realistic Robin shop VRA tracing. The digital avatar Diane.

Oh. Almost perfect. Just that uncanny shimmer in the eyes. Jensen's vision, merge graphics and AI to speed whole universes into existence. Whoa. But AI safety concerns, seemingly dismissed by executives. Catanzaro's electricity analogy, Dierks on AI improving art, Jensen's hot dog example. Brian Catanzaro is the exception, willing to discuss implications. References Arthur C. Clarke, AI as a new phase of evolution. The chapter ends noting a fundamental constraint, power consumption. Even spaceships need fuel. Chapter 3019, power, naturalization.

the massive energy cost of AI. Starts with that lineman in Virginia upgrading power lines for data centers. Dangerous work, relentless demand. AI tasks use way more energy than traditional computing. A GPT query versus a Google search. Huge difference.

And replacing human labor with AI across industries is just going to amplify that energy need massively. NVIDIA's own chips need more and more power, raising questions about needing more nuclear or even reviving coal plants. Yikes. The irony, AI helps study climate change, but also contributes to it. Bill Daley is concerned about this. Plus the business side, is this an AI bubble? Huge investments, uncertain returns for many. But the key players, Hwang, Musk, Altman, seem convinced it's transformative despite the risks.

WIT frames AI as this double-edged sword. Incredible potential, staggering energy costs, maybe some overhype. Just mind-blowing numbers. Revenue doubled to $60 billion, 70% profit margins, net income more than the previous 30 years combined. Crazy.

Huang's statements. AI supercomputers are the new power plants. Trillion dollar AI infrastructure boom coming. Stock surges $277 billion in one day. More than Coca-Cola's entire value. Huge pressure on Huang now. Valuation depends on the future. Lots of direct reports, no clear successor named. The culture.

High pressure, but rewarding for the right people. Top workplace awards, high retention. Still, diversity imbalances remain despite national origin diversity. Geopolitics loom large. Israel-Hamas conflict. U.S. restricting chip sales to China, forcing Nvidia to make modified versions.

the Taiwan TSMC vulnerability. Witt reflects on Huang's success, luck, vision, intellect. The world's betting heavily on him continuing to deliver. Chapter 3221, Jensen, the man, the myth, the cult of personality. Employees practically worship him, partly the stock success, partly his visionary leadership. He's built this persona, speeches, letter jacket, that chaotic stream of consciousness communication style. Witt says he contradicts himself while thinking aloud, then lands on these mantras that define NVIDIA's culture.

That anecdote about him roasting the architect's wardrobe before talking about AI shape-shifting architecture. Classic Jensen, apparently. He's not just running a company. He's shaping the fetcher. His 2024 GTC keynote on natural language interfaces reinventing computing.

But there's a human side. Hates public speaking, worries constantly, doesn't know what to do with his money and his current big project. Maybe strawberry, AI generating mathematical proofs, aiming beyond computation towards reason itself. Wit paints him as unique.

thinker, visionary, maybe a bit mad scientist, changing the world with GPUs. Chapter 33, 22, The Fear. Now a darker turn. Existential dread among AI pioneers, Yoshua Bengio's experience. Yeah, Bengio's sudden realization in early 2023, rapid AI could be devastating. Worse than nukes, maybe. AI designing pathogens. Chillin'. And Hinton and Sutzkever feel it, too. Hinton left Google to warn about risks. Sutzkever focuses on alignment. The

These aren't lightweights. They're giants in the field. Contrasted with Jan LeCun, who dismisses the fairs, argues AI lacks a will to dominate, humans set the goals. Witt introduces peak doom probability of AI extinction. Estimates vary wildly. LeCun, 0%. Bengio, 50%. Hinton, 10%, 20%. Jensen, you're negative. Even a small risk is huge when the stakes are extinction.

And there's no historical data for this. The pace of development is just so fast. Sutzkever's shift to alignment over just building bigger models is significant. But Silicon Valley seems reluctant about regulation. That VO to California bill, SB 2047, wit ends sharing the narrator's personal sense of this fear, questioning humanity's future if machines surpass us. Deep stuff. Finally, Chapter 3423.

The thinking machine, the thinking machine. That last interview with Jensen. After GTC 2024, conference room, diagrams, Huang looks exhausted. He turns the tables, asks about zero-cost math impacting jobs, criticizes the lack of an answer. Then the negative reaction to the Arthur C. Clarke clip, pale face, angry outbursts about pedestrian questions on job loss. Uses analogies, calculators, agriculture, electricity. Reduce costs, society adapts, new jobs emerge. That's his view.

Witt contrasts this with visiting NVIDIA's supercomputer EOS. The sheer power. Reflects on Jensen's pragmatic builder mindset versus speculative futurism and the fear his team seems to have of him. The book and our deep dive ends with this feeling.

We've glimpsed the future being built by this powerful thinking machine, NVIDIA, steered by Jensen, leaves you awestruck but also uneasy. So wrapping this up,

Some key things really stick out, don't they? Jensen Huang's incredible journey, resilience, vision. Absolutely. And the evolution of AI itself from this niche academic thing to something utterly transforming society. Plus the sheer power and complexity of NVIDIA as a company. Right. And considering everything we've talked about, the speed, the potential, those contrasting views from pioneers like Huang versus Bengio.

It leaves you wondering. It really does. Maybe the final thought for you, our listener, is based on all this, what future for these thinking machines seems most likely or maybe most compelling to you? And where do we as humans fit into that picture? That is definitely the question to ponder. Thank you for joining us on this deep dive through the thinking machine. We hope it's given you a good handle on the key ideas in WIT's really compelling narrative.