AIThe case for more empirical LLM findings on LessWrong
One commenter says LessWrong feels less interesting because it has mostly stuck with pre-GPT-3-style essays rather than shifting toward empirical findings about large language models.
TLDR
A commenter criticizes LessWrong for not pivoting toward more “look what I found” accounts of empirical discoveries with large language models. They say the forum mostly retains its pre-GPT-3 essay style, making it less interesting to them. They also allow that they may be partly responsible: they haven’t written those kinds of posts there themselves.
Combined views
348
2 Sources, first seen 4h ago
The case for more empirical LLM findings on LessWrong
One commenter says LessWrong feels less interesting because it has mostly stuck with pre-GPT-3-style essays rather than shifting toward empirical findings about large language models.
TLDR
A commenter criticizes LessWrong for not pivoting toward more “look what I found” accounts of empirical discoveries with large language models. They say the forum mostly retains its pre-GPT-3 essay style, making it less interesting to them. They also allow that they may be partly responsible: they haven’t written those kinds of posts there themselves.