Leopold Aschenbrenner is an artificial intelligence (AI) researcher. He was part of OpenAI's "Superalignment" team, before he was fired in April 2024 over an alleged information leak. He has published a popular essay called "Situational Awareness" about the emergence of artificial general intelligence and related security risks.
Biography
Aschenbrenner was born in Germany. He did research for the Global Priorities Initiative at Oxford University and co-authored a 2024 working paper with Philip Trammell of Oxford. He also has some experience with the effective altruism movement.
OpenAI
Aschenbrenner joined OpenAI in 2023, on a project called "Superalignment" that researches how potential future superintelligences could be aligned with human values.
In April 2023, a hacker gained access to OpenAI's internal messaging system and stole information, an event that OpenAI kept private. Subsequently, Aschenbrenner wrote a memo to OpenAI's board of directors about the possibility of industrial espionage by Chinese and other foreign entities, arguing that OpenAI's security was insufficient. According to Aschenbrenner, this memo led to tensions between the board and the leadership about security, and he received a warning from human resources. OpenAI later fired him in April 2024 over an alleged information leak, which Aschenbrenner said was about a benign brainstorming document shared to three external researchers for feedback. OpenAI stated that the firing is unrelated to the security memo, whereas Aschenbrenner said that it was made explicit to him at the time that it was a major reason. The "Superalignment" team was dissolved one month later, with the departure from OpenAI of other researchers such as Ilya Sutskever and Jan Leike.
Investment
Aschenbrenner said that he started an investment firm with investors Patrick and John Collison, Daniel Gross, and Nat Friedman.
Situational Awareness essay
Aschenbrenner wrote a 165-page essay named "Situational Awareness: The Decade Ahead". It contains sections that predict the emergence of AGI, imagines a path from AGI to superintelligence describes four risks to humanity, outlines a way for humans to deal with superintelligent machines, and articulates the principles of an "AGI realism". He specifically warns that the United States needs to defend against the use AI technologies by countries such as Russia and China. His analysis is based on future capacity for AI systems to conduct AI research, what a Forbes writer referred to as "recursive self-improvement and runaway superintelligence."
References
- Allen, Mike (2024-06-23). "10 takeaways: AI from now to 2034". Axios. Retrieved 2024-12-27.
- "Ex-OpenAI employee writes AI essay: War with China, resources and robots". heise online. 2024-07-02. Retrieved 2024-12-28.
- Metz, Cade (2024-07-04). "A Hacker Stole OpenAI Secrets, Raising Fears That China Could, Too". The New York Times. Archived from the original on 2024-12-26. Retrieved 2024-12-27.
- Altchek, Ana. "Ex-OpenAI employee speaks out about why he was fired: 'I ruffled some feathers'". Business Insider. Retrieved 2024-12-28.
- "Ex-OpenAI Employee Reveals Reason For Getting Fired, "Security Memo Was..."". NDTV. June 6, 2024. Retrieved 2024-12-28.
- Field, Hayden (2024-05-17). "OpenAI dissolves team focused on long-term AI risks, less than one year after announcing it". CNBC. Retrieved 2024-12-28.
- "Post Script: Patrick Collison the Swiss dictator; McKillen-Bono whiskey lands ex-Bank of Ireland governor". www.businesspost.ie. Retrieved 2024-12-27.
- ^ Naughton, John (2024-06-15). "How's this for a bombshell – the US must make AI its next Manhattan Project". The Observer. ISSN 0029-7712. Retrieved 2024-12-27.
- Toews, Rob (November 5, 2024). "AI That Can Invent AI Is Coming. Buckle Up". Forbes. Retrieved 2024-12-27.