A 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis

A 10-year-old CNN outperformed two frontier LLMs in a skin cancer diagnosis task, correctly identifying 62% of 200 biopsy-proven lesions, while the LLMs correctly identified 42% of the same lesions. The CNN was fine-tuned on 68,011 dermoscopic and clinical images from the ISIC Archive on a single H100 GPU, and was used with its default settings, while the LLMs were given both images and asked to provide a diagnosis, confidence, second choice, and discharge recommendation.