AI FOR BIO HAS A FUZZY API PROBLEM

Summary from the TLDR newsletter

Biology is a uniquely hard application domain for machine learning. Drug discovery doesn't have the kind of clean feedback loops and clean interfaces that make modern ML so powerful elsewhere. Software has clean APIs, and when things break, you can usually trace the failure to a bug and fix the code. Biology's outputs are a lot fuzzier as there are many caveats at every step of the process.