CONCEPT
The AlphaFold Objection
The strongest challenge to the prepared-mind framework — a computational system possessing no Pasteurian preparation produced one of the most consequential scientific results of the twenty-first century.
In 2020, DeepMind's AlphaFold predicted three-dimensional protein structures from amino acid sequences with accuracy matching experimental methods, solving in hours a problem that had resisted structural biology for fifty years. The system had no crystallographic training, no years of laboratory experience, no geological strata of perceptual sensitivity. It possessed a training dataset of roughly 170,000 experimentally determined structures and an architecture designed to learn the sequence-structure relationship. It learned the relationship, applied the learning, and produced what prepared minds had not produced in half a century. If preparation is so essential to scientific achievement, how does one account for this? The book's ninth chapter takes the objection seriously — and concludes it proves something specific and limited: computational pattern detection excels at problems within established frameworks where the answer exists and the search criteria can be defined in advance. It does not prove the capacity to recognize when the framework itself is insufficient.
In The You On AI Field Guide
The protein-folding problem is well-defined