Yesterday’s Nibble on the always-on-the-verge-of-breaking-out Bambara groundnut had me rummaging through the archives. Among dozens of references, I came across a post from almost 15 years ago that included some maps, of genebank accession localities and the distribution of the crop. On a whim, I downloaded the Genesys data and fed it into the max of ChatGPT, asking it to identify gaps in the overall ex situ holdings. For each of the top 10 priority regions, I then asked for a best-bet locality for exploration. ChatGPT obliged with a KML file, which I then looked at in Google Earth.
This is the result.
And here’s close-up on West Africa, because that’s where accessions are densest.
Asked for a justification, this is what the LLM came up with.
Does it make any sense? Well, it’s not exactly where I would have plumped for just eyeballing the data. But it is not complete nonsense. Maybe it was the prompt? Any ideas what that should be to get the best results?
Not that any of this is going to help Bambara groundnut much, I suspect.