Will AI/ML methods take over density functional theory?
Resolves YES if an AI-based or AI-generated density functional technique gets more citations (counted by google scholar) than the paper introducing the most popular non-ML-based density functional, (currently B3LYP), has gotten in the TWO previous years by the end of 2027.
Resolves NO otherwise.
Currently, the paper to beat is the one introducing B3LYP with about 1,500 citations / year
2035 (longer term market) here (slight difference in resolution criterion)
1,000
1.00There's no way, based on academic inertia alone.
If I backed a chemical argument up using B3LYP, that choice is usually taken at face value. If I backed it up with some ML thing, now I have to both justify the main point and justify my decision to use that. Not saying it won't ever take off, but certainly not in the next 3 years.
@troops_h8r the main way I imagine this happening would be that a new, ML based method scales well to large systems that b3lyp is intractable for. Still holding NO, though