A few philosophical reflections on artificial intelligence and human thought (long and rewarding ??

Most of us have heard and been deeply impressed by the recent developments in artificial intelligence, and those who haven't should take a moment to chat with chatgpt.com – you can even do it in Hebrew. Personally, I have spent many hours engaging in conversation, learning, and seeking advice on technical matters and more.

The conclusion is that today, this machine called artificial intelligence sounds incredibly intelligent and knows much more than any human being. However... it is not truly wise, and the inferences it makes are as smart as they may seem, they lack genuine understanding. Nevertheless, I must admit that I recently reached a "compromising" conclusion with the machine on a topic in abstract mathematics (for the enthusiasts – Lie algebras). It took quite a while for it to become clear to me, which is not a small achievement.

But let's make a leap and assume that this matter improves. It will improve to the extent that it can produce scientific discoveries as well. Here, we need to pause for a moment and understand the profound difference between human understanding and the "understanding" of this machine. Let's start with human understanding and allow me to summarize in bullet points:

1.??????The universe is vast and complex – this is something most of us can agree on.

2.??????Humans struggle to understand it, so methods of understanding and cataloging based on mathematics have been developed. The scientific and mathematical formulation represents the pinnacle of human achievements; it enables humans to comprehend complex subjects and organize things coherently (quantum mechanics is a notable example, a theory devoid of any human intuition, but provides incredible accurate predictions). Countless books and articles have been written about the relationship between mathematics and the world, and many debates have taken place (for the enthusiasts – "Is God a mathematician?" by Mario Livio).

3.??????Is the universe reducible to mathematical formulation? It is a question that has no answer except for the deep belief of scientists. I, myself, do not think so, and it is not a popular statement within the scientific community, although it has been discussed extensively in the past. How strange is this statement? When my daughter solves one equation, asks for help, and then struggles with another question, I have to explain the same concept again. There is consistency in the way we solve problems. That is our hope, even regarding more complex problems rooted in various phenomena (mathematical or physical): the idea is that different phenomena can be traced back to the same mathematical formulation. Hilbert, a great mathematician, whom I wish I had a tenth of his successes, formulated it as follows, 'Even if a problem is very difficult to solve, if it is approached with sufficient effort and talent, an elegant [mathematical] proof will be found.'

4.??????But at the beginning of the 20th century, Kurt G?del formulated his incompleteness theorem, which cast a big question mark on this endeavor.

5.??????In summary, we rely on mathematics to understand the world, but this tool has its limitations.

Now let's move on to artificial intelligence, once again using bullet points:

1.??????When I look at a piano, I immediately recognize it as a piano. I don't need a mathematical formulation. In fact, to the extent that I understand this process (I'm not a brain researcher), what we do is identify the piano in various resolutions – in other words, the number of quick comparisons made almost simultaneously (in a method called "convolution").

2.??????Artificial intelligence works somewhat similarly to our visual understanding. It compares large arrays of numbers with each other. Its "understanding" is essentially selecting the "correct" arrays of numbers, a mere scrabble game.

Now let's "return to the future", a few years from now. Let's assume that we will be able to use this machine to analyze large experimental data sets and extract knowledge about the universe. This knowledge will not be encoded mathematically, but rather encoded in a way similar to how we encode visual "piano" information. It will understand the world in a different way using the obscure number systems mentioned above. In other words, it will "understand" in a way that is not accessible to humans - and also in a way that cannot be translated into simple coherent laws because (perhaps...) there is no coherent mathematical formulation of all the laws governing the universe.

If that's the case, where is science heading? Towards an accurate "convolutional" formulation of the universe that is not understood by humans and therefore lacks philosophical value for them but has economic value? Or should we continue attempting to stick to a complex mathematical formulation that is not always fruitful, while the knowledge is already beneath the keyboard in an inaccessible black box?"