Episode 19: Minqi Jiang, UCL, on environment and curriculum design for general RL agents
Generally Intelligent
00:00
Is There a Risk of Overfitting?
Co ye mentioned using dolly and clipper bettings as a way of getting realy good representations. Do you tink there's any risk of overfitting to that, or because this is devilly not the way people learn about these things? Ye, that's quite interesting. I see the motivation for thinking why, you know, things might overfit. But right now it just doesn't seem like we've hit that ceiling in an earth scale data set. At this point, like, the models aren't actually even seeing all the data that there is.
Transcript
Play full episode
