Artificial Intelligence is everywhere : in the apps we use every day, in search engines, in chatbots, in the images on our screens. But how does it work, and where does it come from? What can it really do and, above all, how can it change our lives? In the book It's Not Intelligent, But It Applies (Feltrinelli Editore, 2026, pp. 176, also available as an ebook), Alessandro Aresu and Virginia Volpi describe Artificial Intelligence as a great scientific adventure : from the conference that launched it to ChatGPT, from the first computers to artificial neural networks, from ImageNet cats to video games. They uncover the stories and protagonists, the insights, the mistakes, and the epochal turning points that made the most discussed technological revolution of our time possible.

But the question that comes to mind is naturally inspired by the book's title: is he really not intelligent, but is he applying himself? We asked Virginia Volpi:

How many students have heard their parents at least once after a school meeting repeat the professor's fateful phrase, 'He's smart but he doesn't apply himself'? Well, our book is aimed precisely at them. It's an introductory and informative book, suitable for anyone interested in understanding more about AI.
We've chosen a title that's both witty and serious to explain how this technology works. Is it really 'not intelligent'? AI—you'll discover as you read—is trained on a huge amount of data and relies on a probabilistic system. To be more precise, it relies on so-called Large Language Models. When asked a question, these generative AI models return an answer that is the most likely combination of words. Thanks to a series of discoveries, it was Geoffrey Hinton (and others), the 2024 Nobel Prize winner, who realized that to build AI, it was necessary to try to replicate the human brain using artificial neural networks, or many small artificial neurons that receive data and information and process them, layer by layer, in an increasingly deeper way.

La copertina del libro

Can you give us an example to help us understand?

Consider the acronym ChatGPT: Generative, Pre-trained, Transformer. Generative because it generates text; pre-trained because it's trained on a vast amount of data; Transformer because this is the type of neural network used. And it's precisely thanks to Transformers that AI is able to analyze sentences not word by word, but as a whole, identifying connections between words based on context. Based on what was described above, we can understand that AI isn't intelligent, but it applies itself very well. It works probabilistically, which means it can be wrong; it's trained on a huge amount of data and information, but the data used isn't necessarily all correct or always up-to-date, and, in the worst case, it could even be wrong or contain biases. In short, AI is a useful and very powerful tool, but it should be used with an awareness of its characteristics and limitations.
But when was Artificial Intelligence born?

The story of this invention is longer than we can imagine, and is made up of successes and failures, accelerations and setbacks. This book is a journey through this history, beginning in 1955, when the young mathematician John McCarthy first linked the words 'intelligence' and 'artificial', with the goal of building machines capable of using language and solving problems like humans. Machines that could also learn and improve themselves. To do this, in 1956 he invited some of the most brilliant scholars to Dartmouth College, New Hampshire, for what was effectively the first real conference on Artificial Intelligence. This is the official birth date, but reflections and studies had been ongoing for some time, and leading scholars discussed them at this conference. For McCarthy, intelligence was like a puzzle made up of many pieces, which can be assembled and reassembled. For Ray Solomonoff, intelligence couldn't be based solely on rigid rules, as it often had to deal with uncertainty. And the way to account for this was precisely probability. Claude Shannon, who had worked as a researcher at Bell Labs, the major American research and development laboratories, laid the foundations for machines to communicate with each other and exchange information. Finally, it's worth remembering that a few years earlier, Alan Turing, a mathematician and logician, had managed to decipher the messages generated by Enigma (the machine through which the Nazis sent encrypted codes to prevent the Allies from locating them) and had begun to question the possibility of machines learning.
Should we fear AI...or should we understand it?

We fear what we don't know. It's a human reaction. AI is one of the most transformative technologies of our time, and for this reason it arouses enthusiasm, curiosity, but also concern. The goal of Alessandro Aresu's and my book is precisely this: to raise awareness of Artificial Intelligence so we can better understand and use it. To do this, it's not enough to talk about the tools we use today, like chatbots, but it's necessary to look at the history of AI, the scientific and technological ideas that have enabled its development, its functioning and uses, and the complex industrial chain that supports it. Therefore, the first chapters illustrate the ideas that have enabled its development. Subsequently, we delve into the histories of some of the key players, with their concrete insights. Geoffrey Hinton, winner of the 2024 Nobel Prize in Physics for his fundamental discoveries in neural networks and machine learning; Jensen Huang, founder in 1993 of NVIDIA, one of the world's most valuable companies, the first in history to surpass $4 trillion and then $5 trillion in market capitalization; Fei-Fei Li, a Chinese computer scientist and computer vision pioneer, who in 2009 classified countless images, creating a massive dataset on which machines could be trained to recognize them. In the chapter "The Unbearable Lightness of Data," we then describe the robust and enormous infrastructure needed to enable this technology, which is not as intangible as it seems, starting with the fundamental and cumbersome structures of data centers, which consume energy, water, and land.
Finally, the chapter "Everyone Makes Mistakes" focuses on what scares us most: AI errors and misinformation. Yes, Artificial Intelligence—a bit like a 'stochastic' parrot that repeats words without truly understanding them—can sometimes make mistakes, invent information, and experience outright hallucinations, offering what, based on the data, may seem like the most plausible interpretation, but sometimes isn't correct. But, indeed, first of all, we need to understand it.

What is the connection between AI and geopolitics?

Geopolitics is the discipline that studies how a country's geographic location, resources, infrastructure, and economic strength influence its power, security, and relations with other states. Artificial Intelligence fits into this framework for at least two reasons: its production chain is highly concentrated in a few countries, and it is itself an instrument of power. It is no coincidence that this close connection is at the heart of Alessandro Aresu's book Geopolitics of Artificial Intelligence (Feltrinelli, 2024), which reconstructs the intertwining of technology, business, and US-China competition. The AI supply chain begins with raw materials and energy: copper, silicon, aluminum, gallium, rare earths, whose extraction and, above all, refining is concentrated in a few countries, with China holding a dominant position in several strategic materials. Then we move on to semiconductors: the United States leads in design, with NVIDIA leading the way thanks to GPUs (graphics processing units), but the manufacturing of the most advanced chips depends on Taiwan (TSMC), a crucial geopolitical hub. South Korea also contributes to the supply chain, with Samsung and SK Hynix, which dominates memory production. Finally, there's the cloud: thousands of chips gathered in energy-intensive and connected data centers, transforming physical infrastructure into a global service. The US combines big tech, venture capital, universities, and cloud infrastructure; China responds with manufacturing, the domestic market, and subsidies. The supply chain is concentrated and strategic, which is why, for example, the United States has imposed increasingly stringent controls on exports to China of advanced chips (including NVIDIA's GPUs) and the equipment to produce them, using technology as a genuine foreign policy lever.

How can AI improve our lives?

AI can improve our lives on a wide range of levels: from small, repetitive, and boring daily tasks to scientific breakthroughs in medicine or space exploration. Let's start with the simplest things. Many of our daily bureaucratic tasks could be simplified, like filling out forms or writing emails, freeing us up for more creative pursuits.

Moving on to more complex, crucial, and extraordinary tasks, AI has already produced results in the field of medicine: AlphaFold, for example, is an AI program developed by DeepMind to predict the three-dimensional structure of proteins, enabling a better understanding of diseases like Alzheimer's or Parkinson's, contributing to early diagnosis, and consequently enabling the development of new, more effective drugs. Behind this discovery is another of the protagonists cited in the book, Demis Hassabis, winner of the 2024 Nobel Prize in Chemistry. And finally, there is space and the planets of the constellation, where AI-equipped robots and rovers—think of NASA's Martian rovers—must navigate, analyze the terrain, and make autonomous decisions even when communications with Earth necessarily require significant time differences, sometimes as much as half an hour. AI can therefore enable more ambitious missions, in distant environments hostile to direct human control.

© Riproduzione riservata