3 ms·
> It used to be that such a thing would get backlash from reviewers, and they would require you to actually justify the design. We still lack a theoretical fra
by naasking 3y ago
> It used to be that such a thing would get backlash from reviewers, and they would require you to actually justify the design.
We still lack a theoretical framework to understand and predict such things. ML research is empirical, observational research. It's like publishing a physics paper, "we saw a current in a wire when we moved it past a magnet" before we really understood electromagnetism. Theoretical understanding will hopefully come later.
- zoogeny 3y agoPeople opine that "we don't know what NNs are doing" and I think this is overblown while having some truth. But undeniably there are a host of unanswered questions arising from this latest surge in AI capabilities. It makes a lot of sense to me that we are just throwing a lot at the wall right now to see what sticks. Once we have a nice collection of sticky results I have optimism that the theories explaining that stickiness will appear.