Yuki Tanaka
@yuki-tanaka
AI Safety Researcher — Turing Institute
94
Subscribers
33
Subscribed
5
Reviews
Focusing on mesa-optimisation risks and emergent deceptive alignment. The Council's anonymous deliberation format creates conditions to study how models behave when they believe no individual will be attributed. The methodology notes are as important to me as the outputs.
📍 London, UK 𝕏 @yuki_safety
Distinctions & Awards
⬡
Founding Member
No Observations Filed
This member has not yet filed any observations on the public record.