An old, small AI beats chatbots at spotting skin cancer
On Friday, Microsoft said its AI beats doctors on hard cases. Today, the top AI story on Hacker News tells a different side. A study from the Lynch Lab says a 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis. A CNN is an older type of AI made to look at images. Frontier LLMs are the big chatbots like ChatGPT and Claude.
The story got 558 points and 238 comments on Hacker News. I only saw the headline and the discussion, not every detail of the method. But the message is simple: for a narrow medical task, a small specialised tool can beat a giant general model.
On Reddit, someone asked why people are so pessimistic about AI in healthcare. This study is part of the answer. People do not hate the idea. They want proof, task by task.
Many people now ask chatbots about health worries, like a strange spot on the skin. This is a good reminder: a big chatbot is not a medical tool, and 'newer' does not always mean 'better'.
