Verbosity decreases accuracy in large language models (unite.ai)

Where we end up because of this scares me. Literally going back to caveman speak

See? Already doing it.

Amusingly, talk like caveman being discussed here: https://news.ycombinator.com/item?id=47647455