we use the term "Large language models" because the entire world wide web, library of congress, etc. have produced a truly vast amount of written content such t

we use the term "Large language models" because the entire world wide web, library of congress, etc. have produced a truly vast amount of written content such that LLMs have massively large datasets available for learning. That's what I understand to be the "large" part. We have an unbelievable amount of written content available from a huge amount of datasets (both public domain and more questionably public domain). When this video refers to "large behavioral model", where's the "large" part? Where are they getting a similarly "large" amount of behavioral input data? It looks like they have a big lab with a few dozen people modeling behaviors. that's great but it's not like this number of people could produce as much content as all of digital written content.