Opinions expressed are my own and not endorsed by anyone.
Formerly @ ARC Evals aka METR
lukehmiles
Karma: 609
Work dumber not smarter
Wrong answer bias
The economy is mostly newbs (strat predictions)
We can do better than DoWhatIMean (inextricably kind AI)
Decent plan prize announcement (1 paragraph, $1k)
Decent plan prize winner & highlights
Do yourself a FAVAR: security mindset
Distinguishing misuse is difficult and uncomfortable
Could fabs own AI?
Go flash blinking lights at printed text right now
Series of absurd upgrades in nature’s great search
Inner alignment: what are we pointing at?
Creating Environments to Design and Test Embedded Agents
Any evidence or reason to expect a multiverse / Everett branches?
[Question] When did you orient?
[Question] When should I close the fridge?
Could you state the problem and solution more succinctly?
You know making the omnicide machine really is the fastest way to understand how you could have done it safely.