Also, Dan Hendrycks works at xAI and makes capability benchmarks.
He definitely works mostly on things he considers safety. I don’t think he has done much capability benchmark work recently (though maybe I am wrong, but I figured I would register that the above didn’t match my current beliefs).
Earlier this year
Also, Dan Hendrycks works at xAI and makes capability benchmarks.
He definitely works mostly on things he considers safety. I don’t think he has done much capability benchmark work recently (though maybe I am wrong, but I figured I would register that the above didn’t match my current beliefs).
Earlier this year