Comment by thom

Do you think we’ll crack the principles, or do you think we’ll just have very powerful models without really knowing what makes them clever?