AI expert Rodney Brooks is arguing that we're overestimating OpenAI's large language models and that they're far stupider than we think.