Latest / Minds, Bodies, and Terawatts / Google DeepMind Loses Transformer Inventor and Nobel Laureate to Rivals in 48 Hours

Transcript

Google DeepMind Loses Transformer Inventor and Nobel Laureate to Rivals in 48 Hours · Minds, Bodies, and Terawatts

• 0:00
  [Host] This is Minds, Bodies, and Terawatts, where the people who built the future keep leaving the building. Google lost two of AI's most important researchers to rivals in forty-eight hours.
• 0:11
  [Guest] On Wednesday, Noam Shazeer announced he was leaving Google for OpenAI. The next morning, according to Bloomberg, John Jumper said he was leaving Google DeepMind for Anthropic after nearly nine years.
• 0:23
  [Host] What exactly did these two build?
• 0:26
  [Guest] Shazeer co-wrote the 2017 paper "Attention Is All You Need," which introduced the Transformer, the architecture running inside ChatGPT, Claude, Gemini, and every major AI model on Earth. Jumper shared the 2024 Nobel Prize in Chemistry for AlphaFold, the system that predicted the shape of over 200 million proteins.
• 0:45
  [Host] And Google already lost Shazeer once.
• 0:47
  [Guest] He left in 2021 to co-found Character.AI. Google paid about 2.7 billion dollars in 2024 to bring him back. He stayed less than two years.
• 0:57
  [Host] So what is OpenAI offering that 2.7 billion dollars could not buy?
• 1:01
  [Guest] Pre-IPO equity. Both OpenAI and Anthropic filed IPO paperwork in early June, chasing trillion-dollar valuations. A founding-team stake in a company about to go public at that scale is the most powerful recruiting tool in Silicon Valley, and no salary matches it.
• 1:18
  [Host] Sam Altman seemed to agree. What did he actually say when Shazeer arrived?
• 1:23
  [Guest] He posted that Shazeer was one of the people he had most wanted to work with since the very beginning of OpenAI. That is a CEO about to take his company public telling the world he just landed his most-wanted hire.
• 1:39
  [Host] But Google just shipped Gemini 3.5 at I/O three weeks ago. You can lose an inventor without losing the invention.
• 1:47
  [Guest] True, in the short run. But there is a pattern. David Silver, who led AlphaGo, left for his own startup earlier this year. Google keeps building foundational AI and watching the builders leave to compete against it.
• 2:01
  [Host] Here is what concerns me. The talent is not scattering. It is concentrating at two companies about to lock their structures through trillion-dollar public listings.
• 2:11
  [Guest] And once those IPOs price, the incentives harden. Public companies answer to shareholders every quarter. The people who decide what AI costs and who gets access will be the people who own the equity, not the people who use the technology.
• 2:27
  [Host] Patrick Deglon draws that fork in Unscarcity. He calls it the Star Wars path, where a small elite controls the AI and turns abundance into artificial scarcity.
• 2:37
  [Guest] And right now that path is being paved in hiring emails and S-1 filings. The Transformer paper was written inside Google. The Transformer era is being built outside it. For more on the IPO wave and who ends up controlling AI, head to unscarcity dot ai. The book is on Amazon, the audiobook is on Spotify, and this has been Minds, Bodies, and Terawatts.