On Artificial Intelligence

When people first thought whether the artificial intelligence can exceed that of humans, they tried to come up with algorithms to test the computers. One such notable test was the famous Turing test where the human interrogator interacts with a computer and a human without knowing which one is which. If the interrogator could not distinguish which one is the computer, then the computer succeeds the test. Although it sounds like a good test to see if a computer can think as us, it is not exactly fair to the computer as Roger Penrose puts it well in his book The Emperor's New Mind. This is because if we were to reverse the positions, the human would have a hard time to convince the interrogator that she is a computer (Just ask a very complicated calculation question and humans concede.). In a similar manner, why would we assess the abilities of the computer by comparing them to our abilities? Because clearly for a computer to convince a human that it is human, it has to act 'stupider'. While this might be possible (one can argue that today's Large Language Models (LLMs) had already passed this test), would it be a good assessment? Does that mean a computer has a categorically superior mind just because it can imitate the minds of the humans? The kind of intelligence that is only indistinguishable from an human, albeit impressive, would have little use to advance science and technology to the new shores that we could never dream of.

A better assessment could be to judge the computer from its abilities to solve problems as solving problems has always the main driver of the life and civilization. All the progress that has been made on earth can be traced back to a problem and later its solution. One can argue that the speech in humans is an evolutionary solution to solving the problem of coordination or the multi cell organisms is the solution to the problem of efficient utilization of energy. Therefore, whether conscious or not, as long as the computers solve problems better than humans, it shall be deemed more intelligent than humans. When the computer algorithms reveal physical laws that were unknown to us, finish long-standing mathematical proofs, find out new patterns or rules that govern the societies and the universe, solve problems that we could not managed to conjure up a solution, isn't it already something more intelligent than us? Let's imagine an hypothetical supercomputer creates the blueprints to the first successfully built fusion reactor in the world (An enduring obstacle in front of the clean, cheap, omnipresent energy for life on earth). Whether the computer understands what it did will no longer be relevant as the main objective of solving the problem would be completed. However, once can argue that this kind of a computer is just an advanced version of a good-old abacus and helps us to solve parts of a problem that were identified by us, humans. This part is important because a calculator can make calculations that is well-beyond our ability but it would not commence the calculations by itself as it does not have any 'urge' or objective to calculate. Similarly, a computer would not find the solution to the fusion problem unless explicitly instructed by a human which makes the human the subject who controls a powerful tool which in turn makes itself powerful.

Let us move one step forward. Imagine that the reason the computer came up with the solution to the fusion problem was the following. The computer determined that its purpose was to solve problems, solving problems required lots of calculations and this in turn needed electricity. Then the computer identified the problem that prevented it from acquiring more electricity, and it solved the problem: The problem of creating a useable fusion reactor! Here, we did not presume anything about the computer understanding the meaning of any of these things. The computer just determined its objective (solving problems) and the obstacles in front of achieving it (lack of abundant electricity), and came up with a solution (the blueprint for the first fusion reactor). Could there be some similarities here with the aforementioned problem-solution pairs? The humans did not understand the meaning of communication and speech at first but they allowed humans to coordinate and dominate the nature. Similarly, the multicellular organism did not understand what it means to be multicellular but it allowed them to specialise cells, work more efficiently and consequently employ more energy. In the light of all this, should we really expect the computers to understand the meaning of things to deem them intelligent or are we ready to recognise their increasingly good problem solving ability commensurate to our intelligence?

Özgün Barış Kır — November 2025, Evanston / Illinois