These are the medical AI studies I found useful, strange, overhyped, or worth watching. I care most about what changes once the software meets an actual patient.
















Scroll sideways, then hover or tap any dot to open the story.
Arthur Samuel at IBM creates a checkers-playing program and coins the term "machine learning." The first computer that learned from experience rather than explicit instructions.
Stanford's MYCIN used 600 rules to diagnose bacterial infections as well as expert physicians. It proved AI could reason like a doctor, even if it never saw clinical use.
Researchers first applied neural networks to medical images, teaching computers to spot abnormalities in X-rays - the foundation of modern radiology AI.
Geoffrey Hinton's team wins ImageNet using deep learning, slashing image recognition errors. Within years the same tech reshaped medical imaging.
Google DeepMind's AI diagnosed eye disease from retinal scans as accurately as world-leading ophthalmologists - the first clear match to specialist doctors.
A Stanford algorithm trained on 130,000 skin images matched board-certified dermatologists at diagnosing skin cancer. Published in Nature.
DeepMind cracked a 50-year-old mystery - predicting 3D protein shapes from amino acid sequences. A revolution for drug discovery.
The FDA approved the first AI that can diagnose a condition - diabetic retinopathy - entirely on its own, without a doctor reviewing the result.
GPT-4 passed the US Medical Licensing Exam at or above the passing threshold. Physicians began debating AI's role in clinical reasoning.
The first drug molecules designed entirely by AI entered Phase 1 human trials, led by Insilico Medicine and Recursion Pharmaceuticals.
I'm Jack Nassiri, a Loyola High School student who reads medical and training research because the claims are usually more interesting once you open the paper.