antonleicht.me
Evals are crucial for safety-focused AI policy - but four structural problems threaten their future effectiveness as policy instruments. 🧵
Evals are crucial for safety-focused AI policy - but four structural problems threaten their future effectiveness as policy instruments. 🧵
The first episode of frontier AI policy ends and makes way to a different kind of debate. I look at where we are and find precarious achievements, high-profile failures and a difficult political environment ahead.
www.techpolicy.press/the-end-of-t...
The first episode of frontier AI policy ends and makes way to a different kind of debate. I look at where we are and find precarious achievements, high-profile failures and a difficult political environment ahead.
www.techpolicy.press/the-end-of-t...