Robert Leaman
@robertleaman.bsky.social
Staff scientist in biomedical natural language processing / text mining. Creating a literature resource on Long Covid. Views do not represent my employer.
Yes, both as training data and as a source of high(-ish) quality documents for RAG to reduce hallucination. But now that we can generate large quantities of "almost correct" content, the real value might be preserving how we learned something in the first place: the evidence.
Will definitely check out Claude with pubmed integration. This might actually help Pubmed survive... i.e. if AI projects see it as important training data.
Today @anthropic.com released PubMed integration for Claude. No hallucinations. Just real science, real data. As a beta tester, this has been game changing— a supercharged research tool. Here are 6 prompts that will transform how you search the literature. A 🧵
www.anthropic.com/news/claude-...
www.anthropic.com/news/claude-...
October 20, 2025 at 7:32 PM
Yes, both as training data and as a source of high(-ish) quality documents for RAG to reduce hallucination. But now that we can generate large quantities of "almost correct" content, the real value might be preserving how we learned something in the first place: the evidence.
Reposted by Robert Leaman
Sounds about right.
November 16, 2024 at 8:15 PM
Sounds about right.