Now published in open access! Your one-stop shop for the philosophy of language models. It's the spiritual descendant of our two-part preprint from 2024, fully updated. This should be particularly useful for anyone looking for an entry point into this rapidly growing field.
We acknowledge that manual coders and engineers, including C, Haskell, and Java speaking peoples, have stewarded through generations the compilers and datasets of what is now Claude Code. We honor and respect the enduring relationship that exists between these peoples and nations and this LLM.
“Dimicillin” isn’t real. We made it up. Yet many LLMs still call it an antibiotic.
Across 9 models and 653 drugs, we find that drug-name affixes alone can drive pharmacological reasoning. Models often rely on morphology over facts.
We trace this shortcut from behavior to mechanism. 🧵
vincentcarchidi.substack.com/p/what-langu...
Academic writing should be addressed neither to the public nor to other academics. It should be addressed to a single interlocutor with whom you have a petty rivalry and whose work you are trying to discredit
Open call: has anyone managed to make significant use of chatbot LLMs or agents for open-ended research & analysis in a domain that is *not* SWE/coding-heavy or math, while also intending to make the final work public?
(Interpret "significant use" as you wish - this is anecdotal, no wrong answers)
Post-doctoral Research Associate in Computational Linguistics. King's College London.
Salary: £45,031 to £47,379
www.jobs.ac.uk/job/DRU560/p...
New International Construction Grammar Association (ICGA)! icgaonline.wpcomstaging.com announced at ICCG14 thanks to @adelegoldberg.bsky.social
Sign up forms.gle/gXYDKC6xavkU...
@alexbergs.bsky.social H Boas M Fried, @stefanhartmann.bsky.social Th Hoffmann F Masini B Nir @remivantrijp.bsky.social
In film, "we'll fix it in post" is what you say when something went wrong on set and you don't want to redo it. AI research has made it our entire methodology: train the model, then patch whatever comes out. Our new ICML oral argues this can't be the basis of a science of AI. 🧵