Skip to content

Edward Frenkel on the Limits of Large Language Models

"I wonder if a large language model could actually ever come up with the idea that the square root of negative one is something that is essential or meaningful because if all information you get, if all the knowledge that had been accumulated up to that point, tells you that you cannot have a square root of a negative number . . . could one discover that [simply by accumulating everything that had previously been said]?" Edward Frenkel, Lex Fridman, #370 (57:01)