Curated. I think this post is a great demonstration of what our last curation choice suggested
Interpretability research is sometimes described as neuroscience for ML models. Neuroscience is one approach to understanding how human brains work. But empirical psychology research is another approach. I think more people should engage in the analogous activity for language models: trying to figure out how they work just by looking at their behavior, rather than trying to understand their internals.
I’m not yet convinced this will especially fruitful, but this post and others like, e.g. Testing PaLM prompts on GPT3, feel helpful for intuition building about where things are at. I particularly think they help calibrate me about the current SOTA better than the sample outputs that are published when these models are announced.
Curated. I think this post is a great demonstration of what our last curation choice suggested
I’m not yet convinced this will especially fruitful, but this post and others like, e.g. Testing PaLM prompts on GPT3, feel helpful for intuition building about where things are at. I particularly think they help calibrate me about the current SOTA better than the sample outputs that are published when these models are announced.