DeepMind's AMIE diagnostic AI passes first real-clinic prospective test, published in The Lancet
alan_karthi · x · 2026-10-10
After 2+ years, Alan Karthi and colleagues at Google DeepMind/Google Research published the first prospective real-clinic study of conversational diagnostic AI (AMIE) in The Lancet, in collaboration with BIDMC in Boston.
Karthi likens the program to training a junior physician: knowledge exams (Med-PaLM), simulated consultations (AMIE OSCEs), then clinic practice under close supervision.
Key points:
- Patients chatted with AMIE days before a real visit, with chats watched live by a physician trained in safe-stopping criteria — described as a first for prospective medical AI
- Patient satisfaction was high and attitudes toward AI improved significantly post-encounter (p < 0.001); PCPs found the workflow freed time for collaborative counselling
- Top-K diagnostic accuracy held up without EHR context, matching prior simulated results
- AI management plans were less precise than physicians'; real patients rated empathy far more critically than text-based OSCE raters, exposing a disconnect with scripted expressions of sympathy
More from Research
- MOSS@COLM 2026 spotlights small-scale LLM research with talks by Danqi Chen and Omar Khattab — sewon__min · 2026-10-10
- Data firms need frontier-level training infra to prove data value, says Proximal — AccBalanced · 2026-10-10
- AI agent misses $14M in lease clauses: Surge AI's GDP.xlsx benchmarks agents on real spreadsheets — echen · 2026-10-10
- Calico open-sources Cerberus DNA model in PyTorch: 786kb input, 8361 human tracks — anshulkundaje · 2026-10-10
- New preprint uses state space models for long-sequence genomic prediction, open-sources Cerberus PyTorch models — anshulkundaje · 2026-10-10
- MaCVi maritime computer vision workshop lands at WACV 2027 with 3D sonar and underwater benchmarks — HildeKuehne · 2026-10-10