Re: Retrievability
So they've compressed all the bollocks on the internet to produce a machine that can generate similar bollocks on demand? Hmm. An interesting take on the matter...
Has anyone put serious resources into training an LLM with only input from particular kinds of sources, such as the corpus of a traditional publisher? (The kind that use human editors for quality control.) Does an LLM weaned on scientific papers give accurate (and suitably cited) answers to technical questions?