Dwarkesh Podcast Leopold Aschenbrenner — 2027 AGI, China/US super-intelligence race, & the return of history
1410 snips
Jun 4, 2024 
In a riveting discussion, Leopold Aschenbrenner, a valedictorian from Columbia and ex-member of OpenAI's super alignment team, explores the fierce AI race between the U.S. and China. He warns against the perils of outsourcing AI infrastructure and lays out ambitious predictions of achieving AGI by 2027. Leopold shares insights on the geopolitical implications of AI, the importance of safeguarding technology from espionage, and the lessons we can learn from history to enhance national security in this fast-evolving landscape.
AI Snips
Chapters
Books
Transcript
Episode notes
Future Fund Experience
- Leopold worked at Future Fund, aiming to revolutionize philanthropy, before it collapsed due to SBF's fraud.
- He learned the importance of character assessment, even for successful CEOs.
OpenAI Departure
- Leopold was fired from OpenAI's Superalignment team for alleged leaks and internal disagreements.
- He shared a preparedness document externally and raised security concerns to the board.
OpenAI Drama
- Leopold attributes OpenAI's drama to the cognitive dissonance of believing in AGI but not taking its implications seriously.
- He criticizes their broken commitments on compute and security.
