137 emergent abilities of large language models

Eye-catching title and bold claim, but I clicked through on several of the "examples" that the author cites, and I couldn't find any experiments that actually substantiated the claim of emergent properties. Perhaps I am missing something here.