
Yann LeCun on his Start in Research and Self-Supervised Learning
The Gradient: Perspectives on AI
00:00
Is There a Better Way?
There's a trivial solution of everything is the same. So you don't want that. Let's go to collapse, regular collapse or informational collapse. Basically, the two networks happily ignore the input and just produce constant outputs that are always equal. And there's two ways to fix this. The obvious fix that I came up with in 1992 and 1993 was to basically have negative samples.
Transcript
Play full episode