Intelligence to Find the Correct Needle in the Bale of Straw

                                         Intelligence to Find the Correct Needle in the Bale of Straw

I was testing my former colleagues. Yes, now I removed a heavy burden from my shoulders. Till now, I used to hedge my questions to my colleagues in teaching. I used to create deceptively honest queries, almost as honest as the ones that could have come from a student, ignoring the instances when the mischievous student could have been testing the teacher (I as a student had done this!).

When I was a faculty member in a Deemed-to-be university in its School of Civil Engineering, I asked a colleague a fundamental question on mechanics (among faculty members, mechanics does not go beyond the first year Engineering Mechanics, not even Schaum Series book on the subject!). His answer had it been given by his students would have been wrong, but it was not even wrong coming from him. It was absurd. It was, as a judge of the Indian Supreme Court once said, “non-application of mind.” I did not like that put down but I am now using it! Times do change, don’t they?

Guess what! He wanted to impress me and a couple of our colleagues nearby saying that he got the answer through ChatGPT. “See, I know how to use Chat GPT!”. Yes, he sure did, but could not get the correct answer! It was not that he is not intelligent but that he could not identify stupidity!

This is where the heading of this write-up makes its grand entrance. It is also where I can get out of my cocoon and say a thing or two about Artificial Intelligence. My low level argument against Artificial Intelligence, going beyond semantics, is that if the intelligence is witnessed through what has been devised by humans−remember, AI did not drop on to your lap from Heaven, like Manna did on the wandering Israelites on their journey towards their Promised Land. Therefore, Artificial Intelligence in no more than higher level natural intelligence, like human’s as compared to chimpanzee’s, our closest evolutionary cousin.

Per thermodynamics, there might be millions of configurations in a particular situation but the true one is the one having highest entropy, one of highest disorder, or one which would have zillions of axes of symmetry and there are many other such understandings (though this was a favourite subject of mine, I did not understand the concept when I studied it and do not now, more than five decades after).

Just imagine that into an empty vat you drop a white marble (the correct answer) and thousand black marbles (wrong answers). If you pick a marble randomly it would likely be the black marble, the wrong solution. Yet, I would grant that AI does not do the picking randomly. But, is its pick a studied one, a rational one? Perhaps not. It is this “Perhaps not” that led my colleague in teaching to pick the black marble; something like the Black Swanidea for which Prof. S. Varadan of Courant Institute, New York got the Abel Prize in mathematics.

As how AI is being developed is beyond my pea brain, so I would not venture further. Yet, I claim that what thermodynamics and probability of very large numbers show is that one needs to be intelligent to use Artificial Intelligence! Perhaps OpenAI and all the other such claimants demand more than zillion zillion Watts of energy. They, as things stand, need milliamp (mA) energy moderated by neurons in the human brain.

I am not an unreformed Luddite, but perhaps a reformed yet possibly uninformed one who does search the bale of straw for that intelligent needle, either human or Artificial.

Raghuram Ekambaram