
We have already seen many successful applications of AI that are used to solve well-defined scientific problems (e.g. AlphaFold). Now there is an rapidly emerging trend that AI is being used in a different way: there AI goes beyond solving well-defined scientific problems and starts playing an active role in the scientific research workflow itself -- from literature review and hypothesis generation to experiment design and execution, as well as paper writing and review. These AI are the so-called "AI scientist" or "AI co-scientist". In this slides, I give an overview of the frontier work, of which most has been published in top-tier journals in the past 3 years (2023-2026.05).
