What can we use LLM-Clip for to do historical research?
The most obvious after exploring LLM-Clip looks like taking archives/datasets/etc. and using vibe searches to find parallels and similarities among images for sorting purposes and finding new connections. But what else can we do with the model?
A couple thoughts on this:
If I took a gallery that I scraped from a GeoCities archive of just one neighbourhood, I could use the vibe search and define the neighbourhood in its own words and see just how much similarity and connection comes through. Can that help define community or connection within its context?
If I took a medieval bestiary and added scanned images of only the drawn depictions of animals/creatures, then repeated the process including written language, and asked the model "evil incarnate", which depictions would come up either time?