That's certainly the moral. I think what's changed with the new deep learning paradigm is a willingness to accept that implicit prior, not so much because it's

That's certainly the moral. I think what's changed with the new deep learning paradigm is a willingness to accept that implicit prior, not so much because it's good as because we don't know how to improve it consistently.