I am still an AI sceptic, for whatever it’s worth. Not that I doubt its ability or inevitability but am not yet sold on its necessity. However, my daughter, inspired by my techie sister, already has a chatgpt account.
Offline, I was helping her with Algebra. An answer we arrived at (correctly, I insisted) did not match the answer given in the NIOS textbook – unfortunately, not an infrequent occurence. Fed up with the endless methods of factorising a polynomial and the errors in the book, she decided to have some fun and fed the problem to chatgpt to verify the answer.
Not surprisingly, it gave a wrong answer. She kept prodding it but it kept giving the wrong answer. Finally she keyed in the right answer and chatgpt apologized for the wrong answers and accepted the answer she gave as the right one. I still don’t know if it was bluffing, since, when my sister asked the same question a couple of days later, it continued to give wrong answers.
Yes, per se, nothing exciting about this. It’s still early stages, and, no doubt , the machine will get better. What grabbed my attention was the nature of the mistake. It was not a logical error or a formula error. Chatgpt appears to have made a careless mistake. a^4+b^4 became (a^2)^2 – (b^2)^2 in the subsequent step. I would have thought this is not an error a machine is capable of. Maybe, it is indeed getting closer to human intelligence and has developed the ability to make careless mistakes, which should, like love or hate, be confined to the human realm. It’s indeed scary.
(And now my daughter questions, or rather, proclaims, ‘If AI can make mistakes in Maths, why shouldn’t I?’)
Loud Thoughts 