Librarian Aaron Tay, writing on Medium about the potential benefits of OA for machine learning–based projects:

one other benefit that tends to be overlooked, or at least seldom mentioned in my experience particularly by librarians, is how in an Open Access World, we can use machines to plough through the world’s research literature to look for patterns and even possibly do a synthesis of knowledge, leading to vastly greater effectiveness and efficiency in the way we do research…

It’s a smart, well-informed piece, with a nice gloss on the current status of models fed by academic papers. But “vastly greater effectiveness and efficiency”? Careful what we wish for. Among other things, models trained on existing literature risk generating a new Matthew Effect—a faster dynamic of cumulative advantage for the already visible and well-heeled.

Tay identifies emerging projects that attempt to, for example, aid literature searching and paper summation. He mentions Elicit, SciSpace, and Consensus, all three leveraging OpenAI’s GPT3. They’re fun to play with, but Tay doesn’t say that two of the three (SciSpace and Consensus) are for-profit startups. But that matters a lot: The companies are after profits, not knowledge, and they’re sitting ducks for acquisition by Elsevier et al.