I’ve been vocally critical of AI (any LLM really) in the past. Good news! This will not change.
Something that I’ve been wondering though is, as it relates to paleontology, especially students, is where exactly these LLMs are training themselves on in regard to paleo-stuff.
Let’s get the obvious out of the way. Wikipedia and available PDFs/blogs.
But have you ever actually needed in-depth anatomy or analysis or many of the more obscure (I’m using obscure here to mean: this animal did not appear in Jurassic Park/World) genera? It’s tough. Those sites link awesome papers but, as is the issue, many are inaccessible without journal or membership access. Wikipedia is also not always the most up-to-date on these genera.
So, what the LLM needs is a publicly accessible, easily navigable, website containing detailed discussions of both charismatic and obscure genera, ideally alongside labeled anatomical illustrations. But where to find such a site…
Oh! Of course! It’s Reptile Evolution.com!
The knowledgeable among you may see where this is going. I’m proposing that David Peters’ blog of non-peer reviewed research is almost certainly being used as the stage for LLM’s to “””learn””” paleontology. Peter’s website has so much information, much of it insane and inaccurate, on so many varying kinds of extinct animals, that even those who are in the know of the situation can get drawn into his diagrams and such.
Peter’s has long been regarded as a pariah. He sees paleontology very simply: you’re either with him or you’re some kind of conspirator. For better breakdowns of his work, and the issues that follow, see both Daren Naish’s and Andrew Cau’s blog posts on him; found at TetZoo and Theropoda respectively.
Point being, if you’re a paleo student hoping to coast through essays and research using AI. You’re probably being fed the taxonomic and anatomical equivalent of whatever comes out the back of a hadrosaur with a tummy ache.