Feb 23, 2024 · 1h 20m · allin
E167: Google's Woke AI disaster, Nvidia smashes earnings (again), Groq's LPU breakthrough & more
Chamath Palihapitiya · 22m spoken David Sacks · 16m spoken David Friedberg · 16m spoken Jason Calacanis · 15m spoken
0:00 / 0:00
⌖ your search result is the highlighted band (46:42–47:46). Playback starts there
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In Episode 167 of the All-In Podcast, the hosts break down Nvidia's record-breaking earnings, evaluate Groq's LPU hardware breakthrough, analyze the political and quality control controversy surrounding Google's Gemini AI, and review key geopolitical developments in Ukraine.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 99.6% of the talking time here. How this is scored →
The hosts as informed peer 7.0 Guest teaching 2.3 Guest disagreement 3.1 The hosts pushing back 3.3
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 53:45 Politicized AI Guardrails
Sacks aggressively attacks Google's AI principles, arguing that AI safety has been subverted from preventing rogue superintelligence to enforcing political narrative control.
Hardest push from the hosts ▶ 1:00:50 Rejection of Relativist TruthSacks directly refutes Friedberg's proposal that truth output should be a matter of customizable user preference, demanding that models adhere strictly to objective factual baselines.
Biggest teaching moment ▶ 46:44 Compute Architecture MasterclassChamath systematically educates the room on the physical and algorithmic differences between CPU serial processing, GPU parallel matrix math, and LPU language processing units.
The host holds their own ▶ 1:11:54 Google Turnaround BlueprintChamath demonstrates deep executive expertise by laying out a definitive operational plan to reallocate 50-60% of Google's headcount costs into a $100 billion data licensing strategy.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Title Sequence & Formal Host Introductions | 5 | 1 | 1 | 1 | Sacks lays out how GPUs process vector math for 3D graphics and why that same mathematical framework applies directly to LLMs during data center build-outs. Calacanis provides introductory market context regarding Nvidia's record market cap gain. | |
| AI Infrastructure Build-Out: Cisco Bubble vs. Sustainable Growth | 8 | 3 | 2 | 3 | Friedberg details corporate balance sheet accounting, explaining how cloud providers capitalize GPU purchases and depreciate them over multiple years to avoid immediate P&L hits. Sacks and Chamath compare current AI infrastructure spending ratios against historical Cisco and Intel metrics. | |
| Groq's LPU Breakthrough & Deep Tech Investment Strategy | 9 | 4 | 2 | 2 | Chamath breaks down the architectural shift from CPUs and GPUs to Groq's LPUs, explaining how deterministic scheduling speeds up inference. He also details his investment heuristic regarding technical risk versus breaking the laws of physics. | |
| Gemini Rebranding and Image Generation Controversy | 7 | 2 | 6 | 6 | Sacks sharply criticizes Google's Gemini rollout, arguing political ideology hijacked safety protocols to distort historical facts. Friedberg brings up Google's stated AI principles, leading to pushback on whether guardrails compromised core truthfulness. | |
| Information Retrieval vs. Information Interpretation | 8 | 3 | 5 | 5 | Friedberg contrasts classical search retrieval with generative information interpretation, using IQ search queries as an example. Chamath explains how human feedback reward models (RLHF) force explicit ideological guardrails, while Sacks rejects user-preference relativism for baseline facts. | |
| Fixing Google: Leadership, Training Data, and Competition | 8 | 2 | 5 | 5 | Chamath presents a turnaround strategy for Google involving 50-60% headcount reductions and $100 billion annual spend on proprietary training data licensing. Sacks advocates resetting company culture to core information organizing goals, pushing back against Friedberg's user-customization proposals. | |
| LLM Authority, Political Prompts, and the Future of AI | 6 | 2 | 3 | 3 | Calacanis tests Gemini's live refusal responses on sensitive political prompts. Chamath warns of a risk where well-capitalized firms buy up proprietary training data to gate access to objective information. | |
| Ukraine War Update and Transnistria Escalation | 5 | 1 | 1 | 1 | Sacks outlines recent military developments in Avdiivka and potential political escalation in Moldova's Transnistria enclave. The hosts brief the geopolitical situation before concluding the episode. |
Statements from this episode (22)
Nvidia sets corporate record with $247 billion single-day market cap gain
“Shares were up 15% on Thursday, representing a nearly two hundred fifty billion dollar jump in market cap. So let's just let that sit in for a second. This is the largest single day gain in market cap. Two hundred forty seven billion dollars added in market ca…”
Nvidia reports $22.1B Q4 revenue and 9x net income growth
“Q four revenue, 22.1 billion. That's up 22% quarter over quarter, up 265% year over year. The net income was 12.3 billion, nine X year over year.”
Sacks: Nvidia's valuation is grounded in profits, unlike Cisco in 2000
“One is that if you look at Nvidia's multiples, they're nowhere near where Cisco's were back then. So the market in 1999 and early 2000 was way more bubbly than it is now. So Nvidia's valuation is much more grounded in real revenue, real margins, real profit.”
Palihapitiya: Most AI applications today are toy apps, not production code
“Most of the apps, the overwhelming majority of the apps that we're seeing in AI today are toy apps that are run as proofs of concept and demos and run in a sandbox. It is not production code.”
Palihapitiya: Nvidia stock will trend upward for next 18 to 24 months
“Is it not going to go up in the next 18 to 24 months? Probably is for exactly the reason you said. It is super set up to have a very good meet and beat guidance for the street, which they'll eat up and all of the algorithms that trade the press releases will d…”
Palihapitiya: Social Capital will not fund startups requiring new physics laws
“I don't want to debate the laws of physics when I fund a company.”
Calacanis: Cafe X airport robot kiosks generate $1M in annual revenue
“The two in San Francisco at SFO are making like I think they, the two of them make a million dollars a year. And it's the highest per square footage of any store in an airport”
Sacks: Craft Ventures makes a specific portfolio allocation for Elon Musk
“No, in our portfolio allocation, we say this much early stage, this much late stage, this much Elon.”
Palihapitiya: Syndicated venture deals inherit the dysfunction of co-investing partnerships
“Splitting deals will never generate great outcomes, in my opinion, is that you take on all the baggage and the dysfunction of these other partnerships.”
Calacanis: Google's Gemini AI refused to generate images of white people
“But earlier this week, users on X started noticing that Gemini would not generate images of white people, even when prompted.”
Sacks: Gemini's flaws accurately reflect the political biases of its creators
“This idea that Gemini turned out this way by accident or because they didn't, because they rushed it out. I don't really believe that. I believe that what happened is Gemini accurately reflects the biases of the people who created it.”
Sacks: Google will not remove Gemini's bias, only make it more subtle
“Now, I think what's going to happen now is in light of this, the reaction to the rollout is, do I think they're going to get rid of the bias? No, they're going to make it more subtle. That is what I think is disturbing about it. I mean, they should have this m…”
Calacanis: Open-source AI will win as users reject biased proprietary models
“You know who the big winner is going to be in all of this? Chamath is going to be open source. Like, cause people are just not going to want a model that has all this baked in weird bias, right? They're going to want something that's open source.”
Friedberg: General-purpose AI Q&A services must offer personalization to succeed
“The only way that a company like Google or others that are trying to create a general purpose knowledge Q&A type service are going to be successful is if they enable some degree of personalization where the values and the choice about whether or not I want to …”
Sacks: Elon Musk fired 85 percent of Twitter's workforce after acquisition
“I saw what Elon did at Twitter. He went in and he fired 85% of the employees.”
Sacks: Gemini's image errors reflect Google's bureaucratic corporate culture
“The ridiculous images generated by Gemini aren't an anomaly. They're a self portrait of Google's bureaucratic corporate culture.”
Sacks: Dominant cash cows allow companies to survive terrible corporate cultures
“The bigger your cash cow, the worse your culture can get. Without driving you out of business.”
Palihapitiya: The AI company that successfully makes outputs deterministic will win
“AI at its core is about probabilities. Okay. And so the company that can shrink probabilities into being as deterministic as possible. So where this is the right answer. We'll win.”
Friedberg: The open internet will prevent AI training data monopolies
“I think the open internet has enough data that there isn't going to be a monopoly on information by someone spending money for content from third parties. I think that there's enough in the open internet to give our all, give us all kind of, you know, the secu…”
Calacanis: Gemini refused simple queries about political legal cases
“I just asked it a couple of times to just to list the legal cases against Trump, the legal cases against Hunter Biden, the legal cases against president Biden, and it will not just list them. It just punts on that. It's really fascinating. Then chat GPT is lik…”
Sacks: The Russia-Ukraine war is not a stalemate, Russia is winning
“As I've been saying for a while, it's not a stalemate. The Russians are winning”
Sacks: Transnistria annexation attempt could trigger major Ukraine war escalation
“Because if something like this happens, it could really expand the Ukraine war. The West is going to use this as evidence that Putin wants to invade multiple countries and invade, you know, a bunch of countries in Europe, and this could lead to a major escalat…”