Even the best AI models studied can be fooled by nonsense sentences, showing that “their computations are missing something about the way humans process language.”
Well, yes. AI models don't extract meaning. They parrot statistically likely responses based on words used. They had to research that?