Do you know of studies that actually evaluate use of LLMs in the public sector or other kinds of knowledge work? Teknologirådet in Norway just published a report evaluating their own use of LLMs, based on careful logs from 35 cases. teknologiradet.no/publication/...
A fully funded 2-year postdoc for my ERC project on climate politics, comparative politics and political economy.
Deadline: 29 July 2026, 10am CET
Starting date: Jan 2027 (flexible)
Feel free to reach out next week at EPSS if you’d like to discuss the position
www.sciencespo.fr/centre-etude...
Teknologirådet har gjennom 35 case dokumentert og vurdert bruk av KI i vårt eget arbeid med teknologivurdering, blant annet til kartlegging av eksperter og aktører, horisontskanning, utvikling av scen...
2 Years - Postdoctoral Position Starting January 2027
Also, maybe others could use (or adapt) Teknologirådets methods here. Here are the forms they used to log their use of LLMs over two 6 week periods. These are in the appendix. teknologiradet.no/publication/...
Remember when Google's mission was to organize the world's information? Now it's to deliver millions of wrong answers every hour
Diane Bolet
"The Oumi analysis also found that 56 percent of the correct Gemini 3 answers couldn't be backed up by the sources Google linked. The AI is giving answers whose origins users can't trace." Really puts that "91% accuracy" number in perspective. Just getting some fact right isn't enough.
As it has been for at least a year at this point, remember when Elon trained Grok on supposedly more compute than anybody had ever used before and all it got him was a 4% uplift on the AIME '24 ? Good times
Everybody who's still on the "we just need more compute" hype train needs to reckon with the incredibly meager performance improvements that scale is now delivering
Jill Walker Rettberg
Bit of a long one from over on Mastodon. (I am once again begging people to just make blog posts)
infosec.exchange/@david_chisn...
But sure, recursive self-improvement and artificial general intelligence is definitely right around the corner
Jill Walker Rettberg
Kevin Riggle
Kevin Riggle
Kevin Riggle
Kevin Riggle
Kevin Riggle
Kevin Riggle
"The court classified Google as a direct infringer because the "AI overview" is its own content, not just a list of search results. Google's AI overviews had falsely tied two publishing companies to scams...
[A]t Google's scale, it means millions of wrong answers every hour"
