Tried out the Galactica Large Language model. It’s trained on scientific data and it is supposed to be useful for generating scientific type documents. If you read the caveat at the bottom of the page it mentions this technology is prone to hallucinations, that’s definitely what I found when trying ti find papers on Kantian Language models.
I am thinking this is an interesting path, an aid to creativity and possibly a new way of approaching ideas. Now we just need that Wittgensteinian Computer I was looking to build.
Galactica: Article on hallucinating language models
https://galactica.org/?max_new_tokens=400&prompt=Article+on+hallucinating+language+models+
Muffin Media Empire 