Episode 20: Hattie Zhou, Mila, on supermasks, iterative learning, and fortuitous forgetting
Generally Intelligent
00:00
How Do You Interpret a Scene?
In humans at least, like, the way that we interpret a scene is dependent on kind of the context that we're building in mind. So what we interpret is clearly dependent on some set of friars in our head. Also think humans do long spurious features. For example, phobias are examples of kind of like what's specific spurious features. It's not like a pixel based feature. Just because I think our data diversity is too high in our visual field in order to get like pixel-based information.
Transcript
Play full episode
