science ๐ฌ AI coding blunder exposes science risk
Astrophysicist Paul Sutter shared a stressful story about his presentation in February. He was presenting a new void-spotting algorithm designed for galaxy surveys, which was partly coded with AI. Just ten minutes in, a colleague pointed out that the algorithm's handling of survey edges was fundamentally wrong. This incident highlights how AI tools can introduce subtle but serious errors, even in top academic circles. Sutter now vows to immediately distrust any output from AI to ensure accuracy.