Large language models seem startlingly intelligent. But what’s really happening under the hood?| The New Yorker
