Why AI Hasn't Cured Anything yet [video]
The video explores why artificial intelligence, despite its rapid advances, has failed to deliver major breakthroughs in medicine like curing diseases. It discusses the gap between AI's theoretical potential and real-world clinical applications, highlighting issues like data limitations, regulatory hurdles, and the complexity of biological systems.
Background
- Despite the explosion of AI tools like AlphaFold and drug-discovery models, no major drug or therapy discovered primarily through AI has yet reached patients. The hype cycle has outpaced the clinical reality.
- Drug development is extremely slow and expensive (often 10+ years and billions of dollars). AI can speed up early-stage prediction (e.g., protein folding, molecule screening), but the hardest bottlenecks remain: clinical trials, manufacturing, regulation, and proving safety in humans.
- Many AI-discovered candidates have failed in trials or never made it past early testing. The field is still in a "show me the evidence" phase, not a "cure delivered" phase.
- The video explores why biology is harder than chess or language: biological data is sparse, noisy, and high-dimensional, and the "rules" aren't fixed like a board game or grammar.
- Key context: AlphaFold (protein structure prediction) is often cited as AI's biggest biology win, but it predicts structure, not function or disease — useful, not a cure.