Making poorly described data FAIR-er using GenAI A small experiment: can ChatGPT-4 improve the metadata, and therefore the FAIR-ness, of badly described datasets already published openly?