A decade-old CNN reportedly beats frontier LLMs at diagnosing skin cancer

A researcher's analysis, shared on Hacker News, claims that a convolutional neural network first developed about ten years ago performs better than current frontier large language models at diagnosing skin cancer. The post links to a write-up hosted on the Lynch Lab website detailing the comparison. The source material provided is limited to the link and discussion metadata, so the specific datasets, metrics, and methodology are not described here. The Hacker News thread has received only a small amount of engagement so far, with no comments at the time of posting.