Meta's scorched earth approach to AI, Tesla's future, TikTok bill, FTC bans noncompetes, wealth tax
精读 · 按主题梳理核心观点、论证与案例。
§01Tipping the maître d' for a table
The episode opens with a reported reservation flipper, a young entrepreneur said to have sold restaurant reservations for as much as a thousand dollars each, about $70,000 in total. A host says many people have had the app idea but nobody has executed it well.
One host describes his technique: fold a $20 or $50 bill into your palm, slide it across the maître d' stand, and apologize for forgetting to book, asking for help if a table frees up. He says it works almost every time and that he now tips $50 where he once tipped $20.
99 times out of 100, it works.
The others dispute the label. One host calls the payment a pre-tip for expected good service; others call it a bribe that gives plausible deniability and lets him break the line and take someone else's spot. The discussion is anecdote, with no data on how often the tactic works.
§02Llama 3 and the scorched-earth thesis
Meta's Llama 3 is the story of the week. A host says it reached the top two on the Hugging Face leaderboard within days, and that Meta is also putting its AI assistant's search box into Instagram, Facebook and WhatsApp and open-sourcing the Quest operating system. Meta's stock fell as much as 16% after first-quarter results, when Zuckerberg said he would spend heavily on infrastructure.
Chamath argues that Meta is scorching the earth: giving models away destroys the economics of rivals, while Meta's own money still comes from monetizing its apps. He replays a clip from 16 months earlier in which he predicted big companies would do this.
the best way to do that is to scorch the earth with these models
He says Groq deployed Llama within 14 hours for more than 100,000 developers, so foundation models hold little economic value and closed models lose if they cannot train and deploy quickly. Friedberg compares it with Google's open-sourcing of Android, and says open source limits competition because venture capital will no longer put half a billion dollars into a foundation-model company.
Friedberg adds that the capital-spending question is hard to judge because outsiders do not know where Meta is spending, and credits founder voting control for letting Zuckerberg and Musk make hard decisions that a committee would resist.