Written by Paolo Azzoni / INSIDE Industry Association / Published on June 24, 2026

I did not want to write another article about artificial intelligence.

Really.

The world certainly does not suffer from a shortage of AI articles: every week brings a new breakthrough, a new prediction, a new existential warning, a new startup promising to change the world, and a new expert explaining why the previous expert was wrong, …

… yet here I am.

Apparently, AI has become stronger than my editorial discipline. So consider this article a mild surrender … not to artificial intelligence itself, but to the growing feeling that something profound is happening around us … something that goes beyond technology, beyond hardware and software, beyond business models.

This time no technical analysis, nor a market report, nor an attempt to predict the future … instead, I found myself collecting fragments, events, statements, announcements and conversations that is taking time to mentally
digest … therefore consider this article as a modest stream of consciousness … less James Joyce than an engineer trying to make sense of a world that appears to be changing faster than his ability to organize his thoughts
about it.

Taken individually, they may seem unrelated.

Taken together, they tell a more interesting story.

A story about intelligence becoming a resource, a service, an infrastructure, an element of power and perhaps even an
environment … a story about how, while trying to build increasingly capable machines, we may be redefining our own place in the world.

And perhaps that is the real reason why AI continues to fascinate and disturb me at the
same time …

… not because of what machines are becoming …

… but because of what their evolution forces us to ask about ourselves.

The end of human “exceptionalism”?
Last year, Sam Altman made a remark that caught my attention1: “My kids will never be smarter than AI. They will grow up vastly more capable than we grew up, and able to do things that we cannot imagine.”

Whether you agree with the statement is almost secondary … what fascinated me was something else entirely: the fact that such a sentence can now be said publicly, by one of the most influential technology leaders on the planet, without sounding completely absurd.

For centuries humanity has operated under an implicit assumption: we may not have been the strongest species, nor the fastest, nor the most resilient … but we were and are unquestionably the species with the highest cognitive capabilities.

… intelligence was our comparative advantage …

… now, for the first time, we are seriously contemplating a future in which this assumption may no longer hold …

… this does not necessarily mean that humans cognitive capabilities will decline … it simply means that the scale of comparison changes …

… and perhaps this is what makes the current AI debate so emotionally charged … the discussion is not really about benchmarks, models or computational power … it is about identity.

What happens when intelligence ceases to be the defining characteristic that distinguishes us from everything else?

The answer is still unclear to me … but the question alone is enough to make many people (including me) uncomfortable.

When intelligence becomes a utility
Another statement from Altman caught my attention2: he suggested that intelligence
may eventually become a utility, much like electricity or water.

At first glance, the analogy sounds appealing: like electricity democratized physical power, AI could democratize cognitive power …

… need more computing? Buy it …

… need more storage? Buy it …

… need more intelligence? Buy it …

… simple, isn’t it? …

… or perhaps not.

Utilities have always been more than technical infrastructures. They are also systems of economic and political power.
Whoever controls the electrical grid, the telecommunications network or the water supply inevitably acquires significant
influence over society.

This is why the idea of intelligence becoming a utility deserves closer attention: for the first time in history, we are witnessing the emergence of a market where reasoning itself becomes a consumable resource …

… more analysis? More tokens …

… more autonomy? A higher-tier subscription plan …

… more sophisticated agents? Premium service …

… in a sense, we are moving from software as a service to intelligence as a service …

… and the implications are enormous …

… because if intelligence becomes an infrastructure, the next question is inevitable: who owns the infrastructure?

Who controls intelligence?
Once intelligence becomes a resource, ownership becomes unavoidable.

For decades we have debated who controls oil, energy, telecommunications networks and social media (we are currently living a good example of this international “discussion” ) … artificial intelligence introduces a similar question, but on an entirely different scale.

Some politicians have already started asking whether AI companies should remain purely private entities. Their argument is surprisingly simple: if these systems are trained on the collective intellectual output of humanity (for example on our books, articles, code, scientific papers, pictures and conversations), should the benefits belong exclusively to a handful of corporations?

I am not sure I know the answer …

… what interests me is the fact that the question is no longer considered absurd.

Only a few years ago AI was seen as an advanced software industry … today it is increasingly viewed as strategic infrastructure

… and history teaches us that strategic infrastructure rarely remains outside political debates for long.

Perhaps the next great technological competition will not concern who builds the best models …

… it may concern who controls them … corporations … governments … international institutions … or perhaps some combination of all three.

The debate has barely started, but it already feels larger than technology … recent events suggest it is moving from theory to practice much faster than many expected.

Some advanced AI systems are now considered sufficiently powerful that governments have begun treating them as
strategic assets (for example Mythos and Fable 5 models). Access restrictions, export controls and national-security concerns, once primarily associated with strategic assets such as semiconductors, energy infrastructure, telecommunications, defence systems, etc., are increasingly entering the AI conversation.

A few years ago it would have sounded bizarre to compare a language model with sensitive geopolitical infrastructure …

… today the comparison has a central position in political discussions.

Even more interesting is the emergence of a new dilemma inside the industry itself: for decades technology companies competed by releasing increasingly powerful products … now some AI developers openly discuss whether certain capabilities should be delayed, restricted or selectively deployed …

… in other words, we may be entering a phase where the key question is no longer

“Can we build it?” …

… the key question becomes “Should everyone have access to it?” …

… historically, societies have worried about technologies becoming too weak …

… AI may be the first widely deployed digital technology that is raising the opposite concern.

AI as a collective actor
For a long time we have been thinking at AI as an assistant … a very capable assistant, certainly … sometimes surprisingly capable … occasionally annoyingly incapable … but still an assistant.

That description may soon become obsolete.

The most interesting recent developments are not about models writing better emails or generating better images … they are about systems that coordinate other systems.

In other words, AI is beginning to resemble an organization rather than a tool.

Imagine assigning a task and having an AI automatically create a team of specialized agents: one performs the work, another verifies it, a third searches for errors, a fourth reviews the results, while dozens of others operate in parallel.

At that point we are no longer talking about a chatbot …

… we are talking about something that behaves remarkably like a company.

The distinction may sound subtle, but I suspect it is one of the most important transitions currently underway …

… we are moving from artificial intelligence as an individual actor to artificial intelligence as a collective actor …

… and history suggests that organizations are usually far more influential than individuals.

Recently, a new startup have been operating with almost no employees3 4, relying almost entirely on AI agents to run its activities … whether this specific example succeeds or fails is not particularly important …

… what matters is that the idea no longer sounds impossible …

… people are worried about machines replacing workers …

… today we may need to consider the possibility of machines becoming organizations.

There is another development that makes this transition even more significant: AI is no longer merely coordinating other AI systems … it is increasingly contributing to the creation of future generations of AI itself.

Today, a substantial share of the code used to develop and improve AI models is now generated by AI5 … researchers
are developing systems that not only write software, but also identify weaknesses, suggest improvements and occasionally
propose entirely new research directions.

This may sound like a technical detail … it is not …

… for thousands of years, technological progress advanced at the speed of human invention: we built tools, and those tools
helped us build better ones.

What is new is that the tool itself is starting to participate in its own improvement.

Researchers have a name for this phenomenon: recursive self-improvement.

The expression sounds rather academic … the implications are not.

The question is no longer whether AI will accelerate human work … the question is whether AI may eventually accelerate technological evolution itself …

… and if that happens, we may discover that the pace of innovation is no longer determined primarily by human limitations.

The mystery inside the machine
One of the most remarkable AI-related images I have seen recently did not come from Silicon Valley …

… it came from the Vatican.

On one side Chris Olah, one of the researchers helping to build some of the most advanced AI systems in the world …
on the other side the Pope, one of the oldest institutions in human history6 7.

At first glance, it looked like an unlikely encounter … in reality, it made perfect sense.

Chris Olah was illustrating a problem that is becoming increasingly difficult to ignore: we are building systems whose internal mechanisms we do not fully understand …

… this is not how most technologies work …

… an aircraft engineer understands why an airplane flies …

… a bridge designer understands why a bridge stands …

… with advanced AI systems, however, the situation is often different: we know how to train them, how to improve them and how to evaluate them … yet we frequently struggle to explain why certain capabilities emerge or why particular behaviours appear.

At one point, Chris Olah openly admitted that some of the patterns observed inside these models remain difficult to interpret … internal structures that seem to reflect human thinking … like introspection, internal emotional states, and behaviors reminiscent of fear, joy, discomfort, and even satisfaction …

… I found that statement both reassuring and disturbing …

… reassuring because intellectual humility is always welcome …

… disturbing because humanity is not accustomed to deploying technologies that increasingly surprise even their creators.

What fascinated me most, however, was the contrast between the two perspectives present in the room.

The technology community asks: “How capable can these systems become?”

The philosophical and religious community ask: “What remains uniquely human if they do?”

These are not opposing questions … they are complementary ones …

… and perhaps the most important AI debate of the coming decade will not concern technology at all …

… it will concern anthropology.

For centuries we have defined ourselves through qualities such as reasoning, creativity, language and problem-solving … … now machines are beginning to demonstrate versions of those same capabilities …

… not identical versions, certainly …

… but sufficiently similar to force uncomfortable reflection.

The challenge is no longer understanding what AI is becoming …

… the challenge is understanding what we are.

There is another reason why this question may soon become more urgent: for most of the current AI revolution, intelligence has remained trapped behind a screen … we interact with chatbots, voices and avatars, but the experience is still fundamentally digital.

What happens when intelligence acquires a body?

Recent advances in humanoid robotics suggest that this moment may arrive sooner than expected8 … the most striking
progress is not necessarily in movement or manipulation, but in expression … robots are beginning to smile, blink, maintain eye contact and reproduce the countless micro-signals that humans unconsciously use to communicate.

This matters because we do not relate to the world through intelligence alone … we relate through presence.

A virtual assistant may be useful … a machine that looks at you while speaking is something else entirely.

The moment AI becomes embodied, the debate may shift from artificial intelligence to artificial empathy … and the most important question may no longer be whether machines understand us, but how quickly we start treating them as if they do.

When Internet stops being about us
Perhaps the strangest thought arrived when I came across reports suggesting that automated systems may now generate a substantial share of Internet traffic.

For decades the web has been designed around humans … humans search … humans click … humans browse … humans buy.

Increasingly, however, machines are starting to perform many of these actions on our behalf: an AI agent can compare products, gather information, book services, monitor markets and evaluate options at a speed no human can match.

This sounds convenient … and often it is … but it also raises an intriguing possibility …

… what happens when the primary users of the internet are no longer people but machines?

At first, nothing obvious changes … the websites remain … the services remain … the interfaces remain.

But underneath, a different ecosystem begins to emerge: machines gathering information for machines … agents negotiating with agents … algorithms generating content consumed primarily by other algorithms.

The internet starts looking less like a public square and more like an automated supply chain.

For years, the “Dead Internet Theory” belonged to the category of ideas that seemed more suited to late-night online
forums than serious discussion. Yet its core prediction was that much of the Internet activity would eventually become artificial.

The original theory was exaggerated and often conspiratorial.

The Internet is not dead … but it may be evolving into something fundamentally different from what its creators originally imagined.

AI is no longer merely a tool operating within our environment …

… it is gradually becoming part of the environment itself.

The physical weight of intelligence
One of the great illusions surrounding AI is that it feels immaterial.

We interact with a chatbot through a clean interface and receive an answer in seconds. The experience almost resembles magic …

… a magic, however, that consumes a surprising amount of electricity, with a significant environmental impact (a concrete
indication that AI is becoming part of the environment).

Behind every elegant interaction lies a vast ecosystem of data centres, communication networks, cooling systems, and power plants … in fact, one could argue that the future of AI may depend as much on energy as on algorithms.

This becomes evident when proposals emerge that seem to belong to science fiction but are already a reality10 11. Consider the idea of orbital AI data centres: satellites equipped with solar panels, computing infrastructure and cooling systems operating directly in space. Until recently, such concepts would have sounded like futuristic fantasies.

The reason is simple: intelligence requires energy, and advanced AI requires a tremendous amount of it. As demand for
computation continues to grow, engineers must increasingly think not only about software, but also about electricity, heat and physical resources.

This reminds us of something important: intelligence is not floating in the cloud. The cloud, as engineers like to joke, is
simply someone else’s computer … and increasingly, a very large one.

The history of civilization has often been shaped by access to energy … if intelligence becomes one of the defining resources of the twenty-first century, energy may become its hidden currency …

… the AI revolution is therefore not only a story about hardware and software. It is also a story about infrastructure, resources and physics.

Even digital dreams must obey the laws of thermodynamics.

The first technological rebellion
One of the most unexpected developments of the past year has not come from laboratories, it has come from university campuses12 .

Recently, in several graduation ceremonies students were booing whenever the topic of artificial intelligence was mentioned.

The scene feels strangely symbolic.

The generation raised with the Internet, social media and smartphones was expected to embrace AI enthusiastically … instead, many young people seem deeply skeptical.

Why?

Partly because previous waves of technology were marketed as empowerment …

… the Internet promised freedom …

… social media promised connection …

… smartphones promised productivity.

AI, by contrast, often arrives wrapped in narratives about automation, replacement and efficiency. In other words, many young people do not hear “new opportunities” …

… they hear “fewer opportunities.”

Whether that perception is justified is almost irrelevant … because perception shapes reality …

… and perhaps for the first time in recent technological history, a generation fears that the future may be built without needing it. That is a powerful sentiment.

Technology revolutions usually generate excitement.

This one increasingly generates anxiety.

It’s worth paying attention to this difference.

The intelligence multiplier
At this stage it would be easy to become excessively pessimistic.

But that would also be misleading because there is another side to this story.

When discussions focus on job shift, we sometimes forget that human intelligence is not only something that can be replaced …

… it can also be amplified.

Consider the experience of leading scientists and mathematicians who increasingly describe AI as a cognitive accelerator … they use it to explore unconventional ideas, navigate vast scientific literatures, identify connections and test hypotheses at speeds that would have been unimaginable only a few years ago.

This perspective fascinates me, also because this acceleration is visible not only in science, but increasingly in creativity itself.

For most of human history, technological progress unfolded slowly enough for societies to absorb it …

… the extraordinary remained extraordinary for years, sometimes for generations …

… today, AI seems to be compressing the very duration of wonder …

… we are amazed by a new capability on Monday and take it for granted by Friday.

As innovation accelerates, their ability to surprise us diminishes. Perhaps this is one of the most subtle consequences of AI: not only does it multiply intelligence, it also shortens the distance between the impossible and the ordinary. The value therefore shifts elsewhere: from execution to imagination, from capability to purpose, from the tool itself to what we choose to do with it.

When everyone has access to the same extraordinary technologies, what ultimately matters is no longer the technology …

… it is the idea behind it.

Perhaps the most important impact of AI will not be replacing average performers …

… perhaps it will be amplifying exceptional ones.

History has often been shaped by a relatively small number of individuals capable of producing transformative discoveries: if these individuals suddenly gain access to dramatically enhanced cognitive tools, the pace of scientific and technological progress could accelerate in ways we struggle to imagine.

The question then becomes not whether AI makes humans obsolete …

… the question becomes whether AI makes some humans extraordinarily more capable and if that happens, the consequences will extend far beyond science.

They will influence economics, healthcare, education, defense, energy and perhaps every domain where human ingenuity matters. The point becomes not artificial intelligence replacing human intelligence …

… but artificial intelligence extending it.

But there is an important caveat: if AI becomes increasingly capable of performing cognitive tasks, the temptation will be to
delegate more and more of our thinking to it … tasks first … decisions later …

… perhaps eventually even learning itself.

Satya Nadella13 recently made an observation that I found particularly insightful: we may be able to delegate work, but we cannot afford to delegate learning.

The distinction is subtle but fundamental.

Every organization possesses two forms of capital … one is technological: models, data, algorithms and computational resources … the other is human: experience, intuition, judgement, relationships and the ability to recognize what truly matters.

The first can be purchased …

… the second must be cultivated.

Perhaps the real competitive advantage of the AI era will not belong to those who possess the most powerful models, but to those who create the fastest learning loops between human and artificial intelligence.

After all, intelligence without direction is merely capability …

… progress still requires purpose …

… and purpose remains stubbornly human.

The scarcity of humanity
Let me return to where we started.

A child born today may indeed grow up in a world where artificial intelligence surpasses human intelligence in many domains … that possibility no longer feels remote.

But I increasingly suspect that intelligence is not the most interesting variable in this story …

… human history has always been shaped by scarcity. At various times we competed for land, energy, capital, information and knowledge (we are currently still doing it) …

Artificial intelligence promises to make certain forms of intelligence abundant …

… and paradoxically, abundance may increase the value of everything that remains scarce:
… judgement …
… wisdom …
… responsibility …
… empathy …
… purpose …

… the ability to decide not only what can be done, but what should be done.

As machines become more capable, our attention naturally shifts toward the qualities that cannot easily be measured in tokens, parameters or benchmark scores.

I began this article by confessing that I did not want to write yet another piece about artificial intelligence …

… maybe that was naive. AI has become impossible to ignore because it is no longer merely a technological phenomenon … it is a social, economic, political and philosophical one.

And while we spend considerable time asking what AI will become, we should not forget a more important question.

What kind of humans do we want to become alongside it?