
Episode 15: Martín Arjovsky, INRIA, on benchmarks for robustness and geometric information theory
Generally Intelligent
00:00
I Think Scale Is Not the Answer
There are things to do, but at least at the moment, there's no good bench mark for anomaly atection. I think scale is part of the answer, but i think it's just not the whole story. The performance out of this zerosho generlization still sacks if you're comper it to anything more directive. And s willams are on the right toa ha, ha, hano owe. So not kernal ones, just liniars willams. It's an underrated think ingto linares wems really interesting. Tol for understanding what is happening. Hen a practical also afew, just, yes. Let me want to look at
Transcript
Play full episode