Conversation

MSE loss for GANs produces a lot of shitty blurry outputs because waffling & being uncertain allows the GAN to achieve lower loss. similarly an AGI trained on the moral objectives of all humanity may produce some shitty blurry amalgam that everyone hates and trap us in it forever
29
242
This Tweet is from an account that no longer exists. Learn more
Replying to and
you're a chinese cybernetician christopher alexander guy so that makes you gofai determinate optimism anyway, not even in the deep learning regime of things
1
2
Replying to and
I think median vs mean estimation might be better. What can half the people get behind vs what’s in the middle of the two extremes?
1