@sennoma It was always snake oil. LLMs cannot read anything. They cannot summarize. What they can do is generate sequences of words that look like a plausible summary to the person who will read them.
Conversation
Notices
-
Embed this notice
Rich Felker (dalias@hachyderm.io)'s status on Wednesday, 01-Apr-2026 23:45:34 JST
Rich Felker
-
Embed this notice
lower-value human capital (sennoma@chaos.social)'s status on Wednesday, 01-Apr-2026 23:45:36 JST
lower-value human capital
Open Access friends, I have a question for you!
In light of the horrors found inside Claude, is there ANY software that you would trust to draft a literature review for you?
Access to the literature for LLMs (or "data mining") was once, and I'm guilty of this, widely touted as one of the as-yet-unrealized benefits of OA.
"No one can read 129,852* p53 papers," I used to say, "but a machine can."
Is this still a possibility, or is it all just snake oil?
*current PubMed number
-
Embed this notice