Eight Things to Know about Large Language Models - Summary
The paper discusses eight potentially surprising claims about large language models (LLMs), including their predictable increase in capability with increasing investment, the unpredictability of specific behaviors, and the lack of reliable techniques for steering their behavior.