A decade-old image model beat top chatbots at diagnosing skin spots.
London dermatologist Magnus Lynch tested GPT-6 Astra and Claude Fable 5.1 on 200 biopsy-confirmed skin lesions. A small image network from 2015, trained for about 20 minutes, named 62% correctly against 42% for both chatbots. But at its default setting it would send home 18 of 30 deadly melanomas, against three or four for chatbots.
Why it mattersThe test is small and not peer-reviewed, but both chatbots often mixed up harmless and dangerous spots. A chatbot's answer about a worrying mole is no substitute for seeing a doctor.
Read in the app — free, in 9 languages
Related stories
Asked who to vote for, Grok chose Republicans 92% of the time.2026-10-25Top AI models write sentences that are harder for people to process.
2026-10-25Mathematicians set out to rewrite OpenAI's 719 AI-made papers.
2026-10-25Claude designed proteins that latch onto deadly scorpion toxins.
2026-10-24OpenAI's AI found a flaw in the code that shrinks web traffic.
2026-10-24