“In an internal test, Lyssen Labs' simulation model correctly identified a more successful conference talk title, while ChatGPT selected the less successful one.”