I'm not the first to say it, but the distinction over whether models do any "actual reasoning" or not seems moot to me. Whether or not they do reasoning, they answer questions with a decent degree of accuracy, and that degree of accuracy is only going up as we feed the models more data. Whether or not they "do actual reasoning" simply won't matter.
They're already superhuman in some regards; I don't think that I could have coded up the solution to that problem in 5 seconds. :)
They're already superhuman in some regards; I don't think that I could have coded up the solution to that problem in 5 seconds. :)