5 comments

[ 2.0 ms ] story [ 19.2 ms ] thread
(comment deleted)
That part was funny:

> Neural networks are useful tools, even comparable to human intuition, but Parkhill and Aspuru-Guzik agree they cannot replace trained chemists. `Chemists [can] make rational choices about molecular structures in the same way a chess grandmaster makes decisions about chess.`

I don't think they have said both phrases as close together as they appeared on the article.

Ha! I do think it goes back to the old romantic thought of chess as a symbol of impregnable notions of human intelligence ...

probably a bit too late for that thought nowadays

This Kaggle competition was 5 years ago, but sounds awfully similar (selective molecular binding energy estimation).

https://www.kaggle.com/c/MerckActivity

It was won by Dahl, Salakhutdinov, Jaitly, Jordan-Squire, and Geoff Hinton using deep learning.