How AI Is Changing Antibiotic Discovery Not From An AI Guy

Speaking from a microbiologist’s perspective rather than as an AI expert, John Stokes grounds the discussion in the practical realities of small‑molecule antibiotic discovery. He explains that effective antibiotics have historically been found through phenotype‑first screening rather than target‑driven approaches—a process that has been slow and constrained by limited chemical libraries. Stokes describes how emerging AI tools now make it possible to explore vastly larger and more diverse regions of chemical space computationally, far beyond what traditional lab screening can reach. While he emphasizes that these methods are still early and imperfect, he highlights how combining machine learning with experimental validation is beginning to accelerate lead discovery and reduce cost, offering a pragmatic path forward in the fight against antibiotic resistance.