What generative artificial intelligence can teach us about machines, intelligence and, perhaps, ourselves.
Artificial intelligence writes texts, generates images, produces software, solves problems and sustains increasingly convincing conversations.
But does all this mean that machines are truly intelligent?
This is the question at the heart of my article “Dentro la mente dell’intelligenza artificiale” (Inside the Mind of Artificial Intelligence), published in the July-August 2026 issue of Sapere.
It is an apparently simple question, but one that contains many others.
What Do We Mean by Intelligence?
One initial difficulty is that we do not have a universally accepted definition of intelligence, even when we are talking about human beings.
Is it the ability to solve problems? To learn? To adapt? To communicate? To understand the world, ourselves and others?
Since Alan Turing’s celebrated work in 1950, computer science has often adopted a pragmatic shortcut: instead of asking whether a machine really thinks, we can ask whether it is capable of behaving as if it were thinking.
With modern Large Language Models, this distinction has become increasingly difficult to perceive.
These systems converse, explain, argue and sometimes even seem to reason. Yet behind this linguistic behaviour we find mathematical models that learn statistical regularities from enormous quantities of data.
They can therefore display intelligent behaviour without necessarily understanding what they are doing.
Acting Without Understanding
This apparent paradox is one of the central themes of the article.
Contemporary artificial intelligence did not emerge by faithfully reproducing the workings of the human mind. Neural networks, statistical learning and language models have instead found effective engineering shortcuts for producing behaviours that resemble some forms of human intelligence.
To use Luciano Floridi’s expression, these systems can agere sine intellegere: act without understanding.
This does not make their achievements any less extraordinary. On the contrary, it may be one of the most interesting aspects of current technological development.
It does, however, require us to be cautious when attributing typically human categories to machines: understanding, intentionality, consciousness and will.
Human Intelligence Is More Than Computation
The comparison becomes even more interesting when we shift our attention from machines to ourselves.
Human intelligence does not seem to be simply the result of a powerful information-processing system contained within the brain. Language, perception, memory, emotions, relationships and interaction with the physical world all contribute to what we call intelligence.
From Howard Gardner’s theory of multiple intelligences to the perspectives of embodied cognition, a long tradition in psychology, philosophy and neuroscience suggests that thought cannot easily be separated from the body and the world in which that body lives.
Artificial intelligence is something different.
It has no biological history, metabolism or lived experience. Its environment consists primarily of data, algorithms and computational infrastructures built by us.
Perhaps, then, the point is not to ask how quickly machines are becoming similar to human beings.
We may instead be witnessing the emergence of forms of intelligence different from our own, with their own characteristics and limitations.
The Question Is Ultimately About Us
This leads to what I consider the most important question.
Perhaps we should stop asking only:
“Is artificial intelligence really intelligent?”
and start asking:
“What are we willing to delegate to it?”
Artificial intelligence systems are increasingly entering decision-making processes in science, healthcare, finance, public administration and access to information.
Algorithms are not neutral objects, however: they incorporate objectives, criteria, data and choices made by human beings.
The challenge is therefore not merely technological.
It is also ethical and, ultimately, anthropological.
Understanding artificial intelligence inevitably means questioning what we mean by intelligence, responsibility, understanding and human autonomy.
And this may be one of the most interesting aspects of the current AI revolution: by trying to understand increasingly sophisticated machines, we may end up understanding ourselves a little better.
The full article, “Dentro la mente dell’intelligenza artificiale”, is freely available on SapereScienza.