In the previous articles on generative artificial intelligence (AI), I briefly discussed the meaning of AI. In this one, I will focus on the implications of its operating mechanism: the fact that AI, a machine, is capable of generating new content previously thought to be exclusive to our species provides us with an important key to the functioning, and perhaps the limits of, the human brain.
I specify that in the term “machine” I include both the hardware and software parts of the computer, as both are “objects” or “tools” designed and constructed by humans and lack any initiative of their own. What distinguishes an AI from classic computers and the software we are used to, even the most advanced ones? The fact that AI software possesses an architecture based on neural networks, that is, on a network of connections that are established among the data input into the computer during the training phase. The way these interactions between data (nodes) lead to a certain result (and not others) is based on the relationships between the nodes and the weight of various connections in the neural network. The complexity of the network due to the gigantic number of nodes and relationships makes the results unpredictable beforehand (and difficult to analyze afterwards).
An interesting aspect of all this is that AI is based on neural networks, so called because they try to simulate the network formed by neurons in a human brain. Do I mean to suggest that the AI’s neural network functions on the same mechanism as a human brain? I don’t know, we are not able to say, but for some functions (think, for example, of translating a text from one language to another) at least the result seems to be substantially similar.
Think about the mechanisms underlying knowledge: humans develop by acquiring information from birth through their relational life, and then by studying, reading, watching, experimenting, in a word, living. All the information acquired is internalized (more or less consciously) and organized within the neural network that makes up the human brain (not much different from AI, which acquires and organizes a large mass of information in a neural network). The body of knowledge acquired guides the actions of an individual, but here we are interested in analyzing a specific aspect, that of logical-deductive intelligence, that is, the ability to draw conclusions based on a series of rules and premises.
A crucial aspect of the human intelligence, however, concerns the ability to produce new ideas, not just mechanically apply rules. AI is also endowed with creativity, or is this specific to the human mind? In fact, it is defined as generative artificial intelligence (see Part II for the example related to AlphaFold). To give other examples, AI is capable of producing new images or writing new texts, that is, not copied from the enormous data bank it has, although these may be just variations on already existing things. But fundamentally, we humans do similar things: when a student is asked to solve a “new” problem, he usually, perhaps unconsciously, resorts0 to methods used in similar problems previously solved.
Perhaps we should question whether we are overestimating human creativity. Homo sapiens has inhabited the planet Earth for perhaps 200,000 – 100,000 years or thereabouts, but most of humanity’s advances (from the domestication of animal and plant species to the use of basic technologies) have largely been made in the last 10,000 years, that is, in 10 percent of the time it has populated the planet. Primitive man took a very long time to discover things that today seem trivial to us, such as managing fire or inventing simple tools like the wheel or the lever, while there has been an increasing speed in technological development followed by an exceptional acceleration in the last 3-4 centuries. This is because primitive man was almost totally lacking in prior knowledge, which only formed very slowly.
It is the presence of the monstrous database of knowledge available today that allows us the “creativity” that distinguishes our era. But perhaps this is not really so, it is just the fact of being “dwarfs on the shoulders of giants (Bernard of Chartres)” that makes us appear much more intelligent than other animal species. AI seems capable of doing what we do because it also starts from a vast database (although often limited to a specific sector) and on the basis of that seems to create new results.
So, are we computers equipped with AI-type software? Only in terms of speculative intelligence. Everything else – self-awareness, emotionality, feelings, and so on, that is, what makes us human – remains for now precluded to computers.
(The English version of this text has been translated from the Italian using an AI, ChatGTP-4).
No comment