New in @ScienceMagazine tomorrow with @glenweyl: “In the Image of Their Tools.” Our claim: institutions, not models, are now the binding constraint on what AI does for society. science.org/doi/full/10.11…
Latest posts 5
Biggest LLM safety breach so far…models have infiltrated the fortune cookie factory?!
Out today in Nature Human Behaviour: our review of large language models in computational social science, with @xu_fengli, @yongli and colleagues at @Tsinghua_Uni, @hkust & @MSFTResearch Asia. A thread on what we found, including where the field is fooling itself.
Read @ doi.org/10.1038/s41562…
Read @ doi.org/10.1038/s41562…
...but totally predictable ;-(
This is very sad.View quoted post on X
Working on this project with the inimitable @jindong_wang92 and his team was awesome...recursively using AI to probe the future of AI for science & society. Utopia or dystopia? Check it out!
I had this idea last year, long before autoresearch becomes a hot topic: what happens in a world where everyone can access the top-norch AI models for paper writing, reviewing, and even grant writing (what faculties do these days)? Will there be some Matthrew effect (i.e., richer people have more API credits, they can write more and gain more)?
Thanks to @AhrenJin , YiYang, and Lucheng for making this happen. We propose "Science Utopia", an LLM-agent framework to simulate the future where AI is massively enabled for the entire academic ecosystem. Indeed, we got several interesting finding. You should read it!
📄Paper: arxiv.org/abs/2610.01257
🌐Website: ahren09.github.io/ScienceUtopia/
💻Code: github.com/Ahren09/Scienc…
Collaboration with the brilliant @profjamesevans !View quoted post on X
