Are we training the language models, or are the language models training us?

→ View original post on X — @_jasonwei