4 ms·
AlphaFold Is the Most Important Achievement in AI–Ever (2021)
- dekhn 4y agoAlphaFold did not solve the protein folding problem. They made a major advance in a subfield, protein structure prediction, and it depends heavily on statistical analysis of a wide range of information (sequence alignments in particular) to make its inferences. It provides little useful info for flexible/unfolded proteins. There's nothing AI about it at all; it's pure ML, we don't completely understand why or how it works, and doesn't answer the interesting scientific questions of why a particular protein folds, nor what flexibility/movement it accomodates in real time. Still one of the greatest achievments in protein engineering of all time. But not AI, and not solving the protein folding problem.
- rvz 4y agoBetter than what I am see with the hype around LLMs. At least this is AI put to good use rather than all the generative bullshit mania.