Nice (old-ish) post by Ryan O'Connor on emergent capabilities in LLMs. I particularly liked this passage:

"It would be useful to have an epistemological hierarchy of tasks, in which salient and measurable abilities are identified and listed as a required antecedent to more complicated tasks which utilize them. Such a hierarchy may help us begin to predict at what scale some emergent abilities may occur, or at least provide an ordering on the set."
1/2
#LLMs #understandingAI #PhilosophyOfAI

Cool little comment paper by Michael Frank on why LLMs and other Large Pre-Trained AI models can be fruitfully studied using methods and techniques coming from developmental psychology:

https://www.nature.com/articles/s44159-023-00211-x

I've been working on a similar approach to understanding such systems, so it's great to see such ideas get their chance in the Nature spotlight!

#understandingAI #ai #DevPsych #llm