Jacob Steinhardt, UC Berkeley: Machine learning safety, alignment and measurement
Generally Intelligent
00:00
How Do You Scale With the Dimension of the Problem?
The basic question there is, how do i learn from this data in a robust way? You use kind of like naively do gradient descent. A lot of those traditional robust estimators scale somewhat poorly with the dimension of the problem. The big upshot as you actually can develop methods that do scale well with the dimension,. That let you get stuff that's almost as good as if you hadn't had any outliers allut that is ro bust against these outliers.
Play episode from 14:15
Transcript
