OT: Google

“A.I. Predicts the Shape of Nearly Every Protein Known to Science”


“DeepMind has released predictions for nearly every protein known to science. On Thursday, the London-based lab, owned by the same parent company as Google, said it had added more than 200 million predictions to an online database freely available to scientists across the globe.”

Any idiot can make predictions, but the AlphaFold predictions from DeepMind are added to ones released before and that were verified to be pretty good.

Interestingly, as Geoff Hinton has noted in his lectures, ‘deep learning’ isn’t all that different from the original backprop ideas. There have been some nice algorithmic advances, but they are more a nice engineering advance as opposed to some fundamental advancement in our understanding of how neural nets (in brains) work. As Hinton has remarked, the interesting accomplishments of ‘AI’ that we keep hearing about arise from the combination of tremendously better computing power than in the 1980’s, along with huge amounts of now online information to use for training, along with the admittedly nice ‘engineering’ advances. But a fundamental new understanding of ‘intelligence’? Not really.