“[Deep learning systems] are challenged in making … the distinction between features that are merely contingent associations … and features that are inherent properties of the category itself.”
— @GaryMarcushttps://medium.com/@GaryMarcus/the-deepest-problem-with-deep-learning-91c5991f5695 …
-
Sure, but it has no problem predicting protein folding: https://deepmind.com/blog/alphafold/ . Go figure!
Deep learning is part of the system; the AlphaFold blog doesn’t make it all that clear how the system as a whole works, how the system manipulates and represents the 3d structures, etc.
-
-
And as Demis said to the Guardian, “We’ve not solved the protein folding problem, this is just a first step.” Best performance on a particular competition ≠ a solution.
-
Well, if that doesn't impress you, perhaps some realtime video generation: https://www.youtube.com/watch?time_continue=1&v=ayPqjPekn7g … . Deep Learning is eating software!
- 11 more replies
New conversation -
-
-
Likely to be using what DeepMind calls "Graph Networks". Here is a review: https://medium.com/intuitionmachine/intuitive-relational-reasoning-for-deep-learning-3ae164f9f5cd … .
- 1 more reply
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.