May 19, 2023 · 1h 28m · allin
E129: Sam Altman plays chess with regulators, AI's "nuclear" potential, big pharma bundling & more
Jason Calacanis · 27m spoken David Sacks · 22m spoken Chamath Palihapitiya · 15m spoken David Friedberg · 11m spoken
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In Episode 129 of the All-In Podcast, hosts Jason Calacanis, Chamath Palihapitiya, David Sacks, and David Friedberg analyze AI regulatory capture, open-source model proliferation, FTC antitrust enforcement, real estate market pressures, and urban public safety policies.
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 99.8% of the talking time here. How this is scored →
The hosts as informed peer 6.7 Guest teaching 3.2 Guest disagreement 4.0 The hosts pushing back 3.9
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 24:48 Chamath accuses panel of being naive on AI threats
Chamath forcefully tells Sacks and Friedberg that they are unbelievably naive and letting ideology override common sense regarding existential AI dangers.
Hardest push from the hosts ▶ 53:30 Friedberg pushes back using FTC filing detailsFriedberg counters Chamath's broad defense of biotech M&A by reading directly from the FTC filing to detail specific anti-competitive drug bundling practices.
Biggest teaching moment ▶ 25:35 Chamath explains parameter weights and polynomial equationsChamath breaks down how transformer architecture and model weights function like multi-variable polynomial equations to explain model leak risks.
The host holds their own ▶ 1:11:40 Sacks counters AI market narrative with real estate metricsSacks uses empirical data from commercial real estate brokers to show that AI tenant demand covers only 1 million of San Francisco's 30 million vacant square feet.
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 |
|---|---|---|---|---|---|---|
| The All-In Podcast Title Sequence | 6 | 2 | 3 | 3 | The hosts discuss Sam Altman's congressional testimony on AI regulation. Sacks outlines how agreeing to congressional narratives serves as regulatory capture to build a moat for incumbents. | |
| Open-Source AI Proliferation & Unenforceable Regulations | 7 | 4 | 4 | 5 | Friedberg details how open-source LLMs like Mosaic ML's MPT-7B run on the edge, making centralized auditing impractical. Sacks and Chamath argue over the validity of comparing AI licensure to the DMV. | |
| Debating Sam Altman's Motives & The 'Gardez' Strategy | 5 | 3 | 4 | 4 | The panel debates whether Altman is pulling up the ladder or building a moat. Sacks corrects Jason's misuse of chess terminology and reframes Altman's strategy as insider positioning. | |
| AI's Existential Threats & Compute Cluster Licensing Debate | 8 | 4 | 6 | 6 | Chamath invokes nuclear weapons analogies and details transformer parameters and weights, advocating for compute cluster KYC. Friedberg counters with compute cost metrics from recent open source runs. | |
| AI Regulation Debate: Safety Concerns vs. Free Market Innovation | 7 | 3 | 5 | 5 | Sacks shares report from the White House AI summit, contrasting short-term licensure bureaucracy with genuine long-term AGI existential risks. | |
| AI Productivity, White-Collar Job Displacement, and Labor Dynamics | 5 | 3 | 4 | 4 | Jason raises concerns over AI white-collar job displacement while Sacks counters that a 30 percent boost in developer productivity expands software creation. | |
| Autonomous AI Risks: Auto-GPT, Cybersecurity, and Nuclear Analogies | 6 | 3 | 5 | 5 | Chamath and Jason discuss Auto-GPT autonomous loops and zero-day cyber exploits. Sacks argues the industry is falling into a panic trap similar to 2016 tech hearings. | |
| Warning Against the 'AI Industrial Complex' & Regulatory Capture | 7 | 3 | 4 | 3 | Sacks warns of an emerging AI industrial complex. The panel reviews Twitter's appointment of Linda Yaccarino as CEO, noting complementary leadership dynamics. | |
| FTC Blocks Amgen-Horizon Acquisition: Biotech M&A vs. Drug Price Bundling | 8 | 5 | 5 | 5 | Chamath argues FTC blocking of Amgen's Horizon acquisition damages early-stage biotech M&A incentives. Friedberg counters by citing the FTC court filing details regarding price bundling leverage. | |
| Live Demo: ChatGPT Web Browsing & OpenAI Mobile App Launch | 7 | 3 | 3 | 3 | Jason demonstrates live ChatGPT browsing features. Sacks recommends antitrust policy focus on anti-competitive bundling tactics like Apple's App Store monopoly rather than M&A bans. | |
| Apple's Mixed-Reality Headset & The AI + AR Vision | 6 | 2 | 4 | 4 | The panel discusses Apple's anticipated AR headset, exploring price points, battery design, and potential AI-driven contextual overlay use cases. | |
| Gallup Housing Survey: Record Low Homebuyer Sentiment & Mortgage Rate Lock-In | 7 | 3 | 2 | 2 | Sacks breaks down Gallup's historic low homebuyer sentiment, explaining how interest rate spikes create mortgage lock-in that freezes residential mobility. | |
| San Francisco Commercial Real Estate Market Crisis | 8 | 4 | 3 | 3 | Sacks presents data from commercial brokers showing San Francisco's 35 percent office vacancy rate, demonstrating that AI startup demand absorbs only a fraction of empty space. | |
| Urban Safety, Self-Defense, and the Daniel Penny Case | 7 | 3 | 4 | 3 | The panel compares self-defense legal standards in recent urban crime incidents in SF and NYC, discussing public perception, media coverage, and prosecutorial discretion. | |
| Addressing Mental Health: Reallocating USPS Resources | 6 | 3 | 4 | 3 | Jason suggests repurposing postal infrastructure into mental health shelters. Sacks analyzes DA election strategies and Open Society Foundations political influence. |
Statements from this episode (39)
Chamath: 13Fs show hedge funds and family offices long Nvidia and AMD
“And it turns out if you look at their 13 Fs, they're all long NVIDIA and AMD.”
Sacks: Altman's regulatory embrace ensures he will shape new AI agency
“And by basically buying into their narrative and agreeing to everything they want, which is to create all these new regulations and a new agency, I think that Sam is pretty much guaranteeing that he'll be one of the people who gets to help shape the new agency…”
Sacks: Existing laws can handle all AI harms raised in Senate hearing
“Every single thing they talked about could be handled through existing law, if they amount to being harmed at all.”
Friedberg: MosaicML published a free, open-source model with large token capacity
“Last week, they published what looks like a pretty good quality model. That has, you know, a very large kind of token input, which means you can do a lot with it. And that model can be downloaded and used by anyone for free.”
Friedberg: Regulating and auditing local LLM deployments will be nearly impossible
“I think that the point of view is just that this is going to be a near impossible task to try and track and approve LLMS and audit servers that are running LLMS and audit apps and audit what's behind the tools that everyday people are using.”
Sacks: AI licensing will replace permissionless innovation with Washington lobbying
“Again, this is going to replace permissionless innovation, which is what has defined the software industry and especially open source with the need to develop some connection or relationship and lobbying in Washington to go get your project approved.”
Calacanis: Sam Altman turned OpenAI closed by stopping open-source releases
“He made open AI, closed AI, By making it not open source.”
Chamath: Warren Buffett equated AI to nuclear weapons
“Two weeks ago, he equated AI to nuclear weapons, which is an incredibly powerful technology whose genie you can't put back in the bottle, whose 99.9% of use cases are generally quite societally positive, but the .1% of use cases destroys humanity.”
Friedberg: MosaicML spent just $200,000 to train its MPT-7B model
“Basically this model is called MPT seven B. That's the name of the AI model, the LLM model that was generated by this group called Mosaic ML. And they spent 200,000 dollars creating this model from scratch.”
Chamath: A malicious actor will print and release hazardous material using AI
“And then what's going to happen is like some chaotic seeking organization is going to print one of these materials and release it into the wild to prove it.”
Chamath: The most knowledgeable AI figures favor regulation over free-market approaches
“And I just think that on a spectrum, it's, this is a unique moment where the most knowledgeable people across every single political spectrum, persuasion, for-profit, nonprofit, Democrat, Republican, right? Elon and Sam, I'll just use those as the two canonica…”
Friedberg: Being anti-regulation on AI has become politically incorrect
“I actually think it's just politically, I think it's politically incorrect to be anti-regulation right now. I think because of what you're saying, well, just give me a second, I think because of what you're saying, Everyone on the left and the right, it's beco…”
Sacks: AGI is the only species-level risk from AI
“I think there's only one species level risk, which is AGI, but that's a long-term risk.”
Calacanis: Putting Kamala Harris in charge of AI policy makes no sense
“To put Kamala Harris in charge of this makes no sense.”
Sacks: Coders represent the biggest labor shortage in the US economy
“And the biggest shortage in our economy is coders.”
Sacks: ChatGPT and large language models are far from AGI
“Look, ChatGPT strings together, and these large language models string together words in really interesting ways, and they give computers the ability to have a natural language interface. That is so far from AGI.”
Sacks: Tech industry will regret validating AI risk narratives
“The tech industry is gonna regret buying into this narrative because the members of the judiciary committee are doing the same playbook they ran back in 2016.”
Calacanis: Linda Yaccarino will run Twitter except product and CTO
“She's gonna take over everything but product and CTO. Elon's gonna stick with that.”
Chamath: Elon Musk leads companies with female top executives across Tesla, Twitter, SpaceX
“He has a female chairman at Tesla, a female CEO at Twitter, and a female president at SpaceX.”
Chamath: Stemcentrx was Silicon Valley's largest biotech exit and a $10B dud
“The biggest outcome in, in biotech private investing in Silicon Valley was this thing called stem centrics and that That thing was a ten billion dollar dud, right?”
Friedberg: Horizon Therapeutics generates $4B annual revenue and $1B-$1.5B EBITDA
“The company that Amgen is trying to buy is called Horizon Therapeutics, which is a company that's doing about four billion in revenue a year about a billion to a billion and a half in EBITDA.”
Friedberg: FTC's antitrust case against Amgen acquiring Horizon is compelling
“I felt like it was a fairly kind of Compelling case. I'm the lawyer on antitrust and monopolistic practices and the Sherman Act, but this was not sorry, let me just say this was not an early stage biotech risky deal that they're trying to block.”
Chamath: FTC blocking pharma M&A will chill early biotech venture funding
“You're going to see a Paul come over the early stage venture financing landscape, because a lot of these guys that are crossover investors that own a lot of these public biotech stocks that also fund the private stocks will change their risk posture if they ca…”
Calacanis: ChatGPT mobile app will 10x users and usage
“If this thing is an app form, that's going to 10 x the number of users, and it's going to 10 x the amount of usage.”
Sacks: Regulators should target anti-competitive tactics instead of blocking mergers
“What I said in the context of tech is that we should focus on the anti-competitive tactics and stop those rather than blocking all mergers.”
Sacks: Venture capital and pharma investments have only two successful exits
“There's only two good outcomes, right? There's IPOs, there's M&A, everything else basically goes, everything else is a zero, it goes bankrupt.”
Chamath: Apple should have acquired Tesla for $4B-$5B
“They should have bought Tesla when they could have the chance for four or five billion dollars.”
Chamath: Tim Cook famously refused to meet with Elon Musk about acquiring Tesla
“Tim Cook famously wouldn't take the meeting. Elon said it. He wouldn't meet with them.”
Sacks: Mortgage rates locking in homeowners will slow interstate migration
“I think over the last few years during COVID, you saw tremendous movement between states. I think that's going to slow down a lot now because people just can't afford to trade houses.”
Sacks: San Francisco office vacancy rate has reached 35 percent
“The SF office market is just a level set is ninety million square feet. They said the vacancy rate is now 35%. So that's over thirty million square feet vacant”
Sacks: AI office demand of 1 million sq ft won't save SF
“AI companies are only, that's only about a million square feet of demand. So, one million out of thirty million is gonna be absorbed by AI, and you know, maybe that number grows over time over the next five, 10 years as we create some really big AI companies, …”
Sacks: Banks will take over SF office towers and face loan write-offs
“All these office towers are eventually going to be owned by the banks, which you have to liquidate them. And then we're going to find out that these loans that they made are going to be written off because the collateral that they thought was blue chip that wa…”
Sacks: Progressive DAs only eagerly prosecute people acting in self-defense
“The only time you can get a Soros DA excited about prosecuting someone is when they act in self-defense or defense of others.”
Calacanis: Prolonged urban lawlessness inevitably triggers vigilante justice
“The fact is, if you do not, if you allow lawlessness for too long a period of time, you get a Bernie Goetz situation.”
Sacks: Jordan Neely had over 40 prior arrests, including attempted kidnapping
“In truth, he hasn't done that in more than a decade, because again, he had some sort of mental break, and since then, he's been arrested over 40 times, including for attempting to kidnap a seven-year-old child.”
Calacanis: Funding nationwide mental health care would ultimately save cities money
“We would save money. We'd save money because a city like San Francisco Could become quite livable or New York if, and then you've got forbid these terrible school shootings. You know, if you avoid even one of them, it's 30 people's lives or 10 people's lives.”
Sacks: Soros funds local DAs to alter laws via prosecutorial discretion
“He went into cities where he doesn't live. And flooded the zone with money to get his preferred candidate elected as DA. Now, the reason he did that was to change the law. And the way that he changed the law was not through legislatures, the way you're suppose…”
Sacks: Soros's Open Society Foundations foments regime change and revolutions worldwide
“Soros has this thing called the open society foundation, which sounds like it's spreading democracy and liberal values, but in fact is fomenting regime change. All over the world. And he's been sponsoring and funding color revolutions all over the world.”