LPC Coffee Hour : Physics, AI, and AI Safety: Why I’m leaving the field

Wednesday 30 Sept 2026, 13:00 → 14:00 America/Chicago

One West, Wilson Hall

The last few years have seen an incredible amount of progress in the capabilities of Large Language Models. They have gone from failing to perform basic middle school math just a few years ago to now solving Millennium prize problems. In the last year, the models have automated more and more of the activities which comprised my own research, and I feel their capabilities improve with each release. There are no signs this rate of improvement will slow down, and plausible scenarios where it speeds up exponentially in the very near future. Meanwhile, recent events such as the OpenAI autonomous hack of HuggingFace have shown that our ability to control the behavior of these models and align them towards human values is not keeping pace with their capabilities improvement. If this gap continues to grow, catastrophic consequences for humanity seem plausible. I am therefore leaving physics imminently to help with AI safety research. I will discuss my decision making process more in depth, what AI safety research looks like, and the growing opportunities that exist for researchers looking to transition.

https://cern.zoom.us/j/61174099955
Login to indico at cern and follow the Zoom link below or contact lpc-pf@fnal.gov for the passcode

GMT20260930-180137_Recording_1790x956.mp4

13:00 → 14:00 Physics, AI, and AI Safety: Why I’m leaving the field

Speaker: Oz Amram (Fermi National Accelerator Lab. (US))

Farewell_Fermilab_AI_Safety_talk.pdf

Recording Link (YouTube)