“in 2022, a team of Google researchers showed that asking language models to generate step-by-step solutions enabled the models to solve problems that had previously seemed beyond their reach” 🆓

Link. “line of research that uses [very esoteric] complexity theory to study the intrinsic capabilities and limitations of language models.”

This article is also the best LLM primer I have read. Quanta is amazing.

PS. People who say we understand LLMs are wrong.

Trust loss during pandemic: “It’s sometimes hard for people to know who is a straight shooter and who is an activist.”

Link. Some physicians active on social media are heavily invested in narratives that are possible but well supported by evidence. They use dramatic publications that are not representative and they often skew results to favor their strong beliefs. They get a lot of attention. The process of knowledge consensus in science is uglier than civilians used to think.