Feb 11, 2023 · 1h 49m · allin
E115: The AI Search Wars: Google vs. Microsoft, Nordstream report, State of the Union
David Friedberg · 26m spoken Jason Calacanis · 24m spoken David Sacks · 24m spoken Chamath Palihapitiya · 20m spoken Neelay Patel · 28s spoken David Guetta · 26s spoken Media Clip Commentator · 14s 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 this episode of the All-In Podcast, the hosts analyze the escalating AI search battle between Microsoft and Google, the legal and economic ramifications of generative AI, Seymour Hersh's report on the Nord Stream pipeline sabotage, and the U.S. fiscal crisis following the State of the Union address.
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 98.7% of the talking time here. How this is scored →
The hosts as informed peer 6.1 Guest teaching 3.2 Guest disagreement 3.4 The hosts pushing back 3.7
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 1:03:24 Sacks and Calacanis clash aggressively over AI copyright and fair use
Sacks mocks Calacanis as playing an amateur lawyer while Calacanis angrily shouts back that he does not care what Sacks wants and insists AI companies stealing content will be stopped by courts.
Hardest push from the hosts ▶ 1:02:32 Sacks and Friedberg reject Calacanis's claims on AI fair use violationsWhen Calacanis asserts that AI text generation is non-transformative rewriting, Sacks and Friedberg completely reject his framing, arguing that AI output is synthesized and transformative.
Biggest teaching moment ▶ 38:20 Friedberg breaks down Search vs LLM query cost economicsFriedberg educates the panel with quantitative back-of-envelope math showing LLM query compute costs 30 cents compared to traditional search at 2.5 cents, explaining the massive financial hurdle Google faces.
The host holds their own ▶ 1:31:23 Friedberg cites Treasury forecast data on US fiscal sustainabilityFriedberg demonstrates strong macro expertise by grounding the State of the Union debate in actual US Treasury data forecasts to demonstrate an inevitable debt default without structural reforms.
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 |
|---|---|---|---|---|---|---|
| Opening Banter: Kids and Phone Etiquette | 2 | 1 | 3 | 4 | The panel opens with lighthearted stories about teenage phone etiquette before transitioning into a debate on the Chinese spy balloon. Calacanis challenges Sacks on his use of the word 'errant' to describe the balloon's trajectory. | |
| Seymour Hersh Report on Nord Stream Pipeline Sabotage | 6 | 3 | 5 | 6 | Sacks presents Seymour Hersh's investigative report alleging US involvement in blowing up the Nord Stream pipeline. Calacanis pushes back forcefully, questioning Hersh's reliance on a single main source and weighing the CIA's official denial. | |
| Media Narrative Machine: Matt Orfalea Montage | 5 | 2 | 2 | 3 | Sacks introduces Matt Orfalea's video compilation showing mainstream news anchors repeating verbatim talking points blaming Russia for Nord Stream. The co-hosts discuss how media narratives are synchronized via surrogate memos and chat groups. | |
| The Military-Industrial Complex & War Incentives | 7 | 4 | 2 | 3 | Friedberg reframes the Nord Stream debate around structural economic and military incentives rather than top-down conspiracies. Palihapitiya and Sacks concur, citing military-industrial lobbying and financial interests driving conflict. | |
| Podcast Banter, 'Alt-Davos', and Retro Photos | 1 | 1 | 1 | 1 | The hosts engage in light banter about podcast appearances, joke about launching an 'Alt-Davos' conference, and tease Palihapitiya over old wardrobe photos. | |
| Microsoft vs. Google: The AI Search Battle Begins | 8 | 7 | 2 | 2 | Friedberg provides a detailed technical and financial analysis comparing classic search information retrieval against LLM query costs. He breaks down how traditional search costs 2.5 cents per query while GPT-3 compute currently costs 30 cents per result. | |
| Microsoft's Strategy to Erode Google's Market Share | 7 | 3 | 2 | 2 | Palihapitiya analyzes Microsoft's strategy with Bing and OpenAI, arguing Nadella is sacrificing short-term margins to force Google to defend its search dominance at higher operational costs. | |
| Sundar's Countermeasure: Doubling Traffic Acquisition Costs | 7 | 4 | 4 | 4 | Palihapitiya proposes Sundar Pichai's ideal countermeasure: doubling Google's Traffic Acquisition Costs (TAC) to lock up default distribution and publisher access. They review a clip of Satya Nadella being pressed on publisher revenue sharing. | |
| Disrupting Google's Ad Model and Search Intent | 7 | 3 | 3 | 4 | Sacks argues that AI direct answers break the traditional ad-click monetization loop. Friedberg counters that commercial intent searches still benefit from transactional conversion links rather than text summaries. | |
| Weaponizing TAC and Exclusive Content Rights | 7 | 2 | 3 | 3 | Palihapitiya elaborates on using TAC aggressively to secure exclusive AI crawling rights from top digital publishers, forcing rivals into costly copyright litigation. | |
| The Open Web, Copyright, and LLM Data Ingestion | 7 | 5 | 5 | 5 | Friedberg argues that reading public web pages to train LLM parameter weights is analogous to human learning and falls under fair use, while Palihapitiya notes that valuable content is increasingly behind walled gardens. | |
| AI Citation Models and Fair Use Arguments | 6 | 4 | 6 | 7 | Calacanis argues LLM answers without attribution violate fair use doctrine by harming original creators' ability to monetize their work, citing Neeva's citation model as a better alternative. | |
| The Debate Over Fair Use and AI Copyright | 5 | 3 | 8 | 8 | A heated argument breaks out as Sacks dismisses Calacanis's legal take on copyright and fair use, insisting AI adoption cannot be halted by copyright claims while Calacanis maintains tech companies will be sued into settlements. | |
| Practical AI Applications: Excel and Everyday Assistants | 7 | 3 | 2 | 2 | Sacks shares a product demo where AI explained complex Excel models and generated formulas instantly. Palihapitiya presents a thought experiment on building a Stripe competitor with only five engineers using AI coding assistants. | |
| Productivity Gains vs. Economic Growth | 7 | 3 | 3 | 3 | Friedberg notes that AI tools unlock massive per-worker productivity gains. Calacanis demonstrates prompt-to-UI tools like Galileo AI and GitHub Copilot's autocompletion capability. | |
| The Future of SaaS and Creative Media Disruption | 7 | 4 | 3 | 4 | Friedberg asks whether easy code generation commoditizes traditional SaaS businesses. Sacks counters that AI will enhance developer productivity and displace legacy non-software industries. | |
| David Guetta's Eminem Deepfake and Synthetic Audio | 5 | 3 | 1 | 1 | Friedberg showcases a clip of producer David Guetta using ChatGPT to write lyrics in Eminem's style and an AI voice synthesizer to generate a viral track. | |
| The New York Times Hit Piece on David Solomon | 6 | 2 | 2 | 3 | Palihapitiya criticizes a New York Times article framing Goldman Sachs CEO David Solomon's DJing side gig as a potential conflict of interest, calling it an unfair hit piece. | |
| DJ Banter, Classical Music, and Summit Planning | 5 | 2 | 4 | 4 | The hosts share banter about playing musical instruments before Calacanis reiterates that online content creators must form a united front to secure licensing revenues from AI developers. | |
| AI Synthesis, Creative Work, and the Long Tail of Content | 6 | 3 | 5 | 5 | Friedberg compares AI model training from online text to classical composers learning from predecessors. Calacanis argues that republishing aggregated reviews harms original media outlets. | |
| AI Copyright Litigation: Getty Images and GitHub Copilot | 6 | 4 | 6 | 6 | Calacanis highlights Getty Images' lawsuit against Stable Diffusion over watermarked training data and the class-action suit against GitHub Copilot, asserting legal precedents will enforce licensing. | |
| State of the Union Analysis: U.S. National Debt and Entitlements | 8 | 5 | 3 | 3 | Friedberg analyzes the US Treasury's long-term fiscal debt projection chart, highlighting how entitlement obligations and interest payments create an unsustainable fiscal trajectory. | |
| Biden's Political Strategy and Congressional Decorum | 7 | 3 | 3 | 3 | Sacks analyzes Biden's State of the Union strategy, arguing Biden used 'small ball' political issues and entitlement defenses to pivot toward his 2024 re-election campaign. | |
| Wealth Tax Debates and Energy Abundance as Economic Solutions | 8 | 3 | 3 | 3 | The panel debates proposed wealth taxes and capital flight, pointing to historical precedent in France and current trends in San Francisco. Friedberg and Palihapitiya conclude that dramatic energy scaling (solar, fusion) is the only path to economic growth out of national debt. |
Statements from this episode (54)
Sacks argues the mainstream media is reflexively hawkish on international affairs
“And it just shows how reflexively hawkish the media is, you know, it's like, Hey, can we just wrap up this war in Ukraine before we start another war with an even more formidable super power?”
Sacks: Hersh report alleging U.S. Nord Stream sabotage is plausible
“I lean towards thinking it's more plausible than not.”
Calacanis: CIA probably has plausible deniability regarding Nord Stream
“The CIA probably has plausible deniability. That's my prediction.”
Sacks predicts German Chancellor Scholz faces political trouble over Ukraine war
“If this war drags on for a long time, I think Schultz might be in some political trouble precisely because he's gone along with the Americans on this.”
Calacanis: Google Bard made a factual error during launch demo
“And Bard answered that it took the first pictures of a planet outside Earth's solar system, which is false”
Friedberg: Google AI improved data center energy efficiency by 40%
“They demonstrated last year that their AI improved data center energy efficiency by 40%.”
Friedberg estimates a ChatGPT query costs 30 cents, 10x traditional search
“A recent estimate on running the GPT three model for chat GPT is that each result takes about 30 cents of compute. So it's about an order of magnitude higher cost to run that search result than it is to do it through a traditional search query.”
Chamath: Specialized silicon will make AI search scalable and affordable
“The ability to run this at scale is going to happen because we're getting better and better at creating silicon that specializes in doing things in a massively parallelized way.”
Palihapitiya predicts AI model output costs will drop 90% in coming years
“The same output today will cost one 10th, as long as you ride the energy and compute curve for the next few years.”
Palihapitiya argues ChatGPT will quickly be commoditized by widespread market competition
“Chat GPT is an incredibly important innovation. But it's an element of a platform who will get quickly commoditized because everybody will compete over time.”
Palihapitiya: Google will generate nearly $100B free cash flow in 2023
“This is a business that this year will do almost a hundred billion dollars of free cashflow.”
Calacanis argues Google has ten times more employees than needed
“And they have 10 times as many employees as they need to run the core business.”
Palihapitiya predicts Google's default search deal with Apple could cost $25 billion
“This year, this renegotiation for that deal could mean that Apple gets paid twenty-five billion dollars for giving away that right to Google.”
Chamath: Google spent around $45 billion on search traffic acquisition in 2022
“Last year they spent, I think, forty-five billion or so, so about 21%.”
Sacks: Generative AI search could disrupt search advertising monetization models
“I think the whole monetization model might change.”
Friedberg: Google historically used bounce back rate as key quality metric
“So at Google, we used to have a key metric was the bounce back rate.”
Friedberg: Google pays Apple $15B annually for default iPhone search placement
“Today, Google pays Apple, fifteen billion dollars a year to be the default Search engine on iPhones on the Safari browser.”
Palihapitiya predicts publishers will block non-paying AI agents from their content
“So this is going to happen for these agents. It's unrealistic to expect that it won't.”
Friedberg: Google AdSense pays out 70 cents per dollar to publishers
“First off, like Google today, just so people know on AdSense is typically paying out 70 cents on every dollar to the publisher.”
Palihapitiya claims transformer LLMs trained on identical data yield identical answers
“I'm saying when you look at transformer architecture today, every LLM that you write on the same corpus of underlying data for training will get to the same answer.”
Calacanis asserts AI web scraping for derivative answers violates fair use
“It's not fair use because in fair use, you have the ability to create derivative works on future platforms. And you are taking this person, the original content owner's ability to exploit that And you are co-opting it and you're doing it at scale. And that is …”
Palihapitiya argues limiting AI tokenization to sentence blocks severely degrades quality
“If you limit How they can tokenize to just being all entire sentences. The product will not be that good. Like the whole idea of these LLMs is that you're running, you know, so many iterations to literally figure out what is the next most best word that comes …”
Friedberg explains LLMs generate non-deterministic statistical inferences, not static discrete outputs
“These are also not deterministic models, and they're not deterministic outputs, meaning that it's not a discrete And specific answer that's going to be repeated every time the model is run. These are statistical models. So they infer what the right answer coul…”
Sacks predicts powerful tech waves like AI won't be stopped legally
“New technology waves that are this powerful, don't get stymied by either chip costs or legal rights issues.”
Calacanis: Companies using AI to steal content will be stopped by legal action
“AI's not gonna be stopped, but companies using AI to steal content will be.”
Sacks: Every software application will feature a voice-based AI interface
“I think in every single application that we use, there's going to be an AI interface and like a, it is probably gonna be voice based where you can just say to it, Hey, I'm trying to accomplish this. Like, how do I do it? Can you just make it happen?”
Chamath: Code generation will be the first area AI excels at with precision
“The ability to write code is going to be the first thing that these guys, that these things do incredibly well with absolute precision.”
Palihapitiya argues lean AI startups will easily disrupt unmoated middleman businesses
“Any middleman business right now that doesn't have its own competitive mode can be competed against. Because now you can take all of those input costs that go into human capital, you can defer that, have a much smaller human capital pool, and push all of that …”
Friedberg predicts one developer using AI matches the output of 20 engineers
“One human can do the coding that 20 humans would do before.”
Calacanis predicts generative AI will enable founders to build MVPs without developers
“These two things are gonna come together where you're gonna be able to build your MVP for your startup by typing in text and then publish it. You're not gonna need a developer for your startup.”
Sacks predicts AI copilots will boost developer productivity, not replace engineers
“I also don't think that, like, you're not going to need to hire engineers because Copilot's just going to do it for you. I think what Copilot will do is make your typical engineer more productive.”
Calacanis predicts generative AI will spark a boom in niche industry-specific SaaS
“I think more startups, more niche startups will make better products And then you'll just have many more folks like making sats for dentists, making sats.”
Palihapitiya warns niche vertical SaaS startups will yield poorer venture capital returns
“It'll be poorer returns.”
Friedberg: AI will completely rewrite the video game industry
“I think the entire video game industry is going to get completely rewritten with AI because you're not going to have a publisher anymore that makes one game that everyone consumes. You're going to have tools that everyone creates and consumes their own game.”
Sacks: AI will preserve actor likenesses in digital databases forever
“No actor ever goes away. You could, there's going to be a database of all of them.”
Calacanis: AI will render video scenes in real time during scriptwriting
“As you write the script, the AI is going to be showing you that scene in real time.”
Friedberg predicts AI compute will collapse movie production costs to under $10,000
“This takes it to a whole nother level where the monopoly That big production houses have because they have the big budget so they can afford to make a big movie. If that cost of making a ten million dollar movie goes to 10,000 or a thousand dollars of compute …”
Palihapitiya asserts technology is fundamentally deflationary, reducing aggregate industry revenue
“I think the point here is that technology is fundamentally deflationary. Here's the next great example where the minute you make something incredible, costs go down, but also frankly revenue and profit dollars go down in the aggregate.”
Chamath: Google should cannibalize its search business before competitors do
“Google should go and they should cannibalize their own business before it is cannibalized for them.”
Friedberg predicts 95% of generic self-help websites will disappear due to AI
“It is more likely than not that many of these quote content publishers that aren't adding very much marginal value are going to go away. That you could see the number of content sites offering self-help advice and how to do this and how to do that. 95% of them…”
JCal: AI platforms will settle with publishers over copyrighted content
“There'll be a settlement where they are going to be able to claim their content.”
Calacanis predicts Stable Diffusion will pay $100 million to Getty Images
“They are going to pay a hundred million dollars or more to Getty Images for stealing their content and allowing it to be republished In a commercial setting.”
Palihapitiya predicts Google losing search market share kills FTC antitrust cases
“Microsoft taking five or 600 basis points of share is the best way to ensure that the FTC has zero credibility in going after Google or anybody else in tech. Those, all of those things I think are DOA.”
Calacanis: AI copyright lawsuits will force mandatory content licensing fees
“These lawsuits are just beginning, and it's gonna result in licensing. Licensing fees.”
Friedberg: Copyright litigation will be transitory and won't slow AI long-term
“This will be a transitory effect, and it won't change the dynamics of where this is going over the long term.”
Friedberg predicts the federal government will bail out underfunded private pensions
“A lot of the private pensions that are going to need to get bailed out with the same federal money, because they're not going to let those things go bankrupt. And that's another trillion plus of liabilities.”
Sacks predicts any Republican proposing entitlement reform will unequivocally lose
“Any smart Republican is going to take entitlement reform off the table because it is a total third rail. And they will lose.”
Friedberg argues 10x energy capacity and 50% cost cuts solve US debt
“If you can get the cost of energy in this country to drop by 50 to 75%, and you can increase energy capacity by 10 to 20 fold, then you have a fighting chance, because you can actually grow the economy out of the problem. And that's really where I am optimisti…”
Chamath: Solar power costs three cents per kilowatt-hour, making it the cheapest energy
“You know what the cheapest source of energy is today? Fusion. It's three cents a kilowatt hour.”
Chamath: Scalable battery storage will unlock zero-cost energy long before nuclear fusion
“The thing that we are lacking right now is not actually the generation capability, which is incredibly cheap, but it's really Scalable storage. And once we figured that out, which is actually the real technical bottleneck to abundant zero cost energy, we'll ha…”
Friedberg claims San Francisco spends an astonishing $18,000 per citizen annually
“It's 18,000 dollars. Per citizen per year. That's how much the city of San Francisco spends.”
Friedberg: U.S. federal budget per capita is $15,000
“The federal government's budget per capita is 15,000 dollars.”
Friedberg: Raising taxes without spending discipline drives away tax base long-term
“Anyway, the point is increasing the tax rate is a great short-term solution, but over the long term, if that budget per citizen isn't brought in line, there is no way to tax the base enough without causing the tax base to leave.”