Living the Future: Artificial Intelligence Between Knowledge, Responsibility, and Care

From artificial intelligence research at Stanford in 1982 to the prospects of 2076. The future will depend not only on the capabilities of machines but on our ability to transform innovation into knowledge, autonomy, and inclusion.

The article originates from the reflections presented at Tecnòpolis 2026, in Bologna, during the meeting "The Possible Horizon: Four Perspectives to Imagine the World," on October 15, 2026.

The future is not written in technology

When, in 1982, I was working at Stanford on artificial intelligence applied to learning, one of the issues that interested me the most concerned the dialogue between people and intelligent systems. How could a machine contribute to the construction of knowledge? In what way could interaction with a computer system modify learning processes?

More than forty years later, those questions are still relevant. The capabilities of systems, computing power, data availability, and the spread of technologies have changed radically. But the fundamental question remains: what kind of relationship do we want to build between human intelligence and artificial intelligence?

It is a question I raised again at Tecnòpolis 2026, in Bologna, during a discussion dedicated to imagining the world of the next fifty years.

The history of technology teaches us that we can glimpse the technical possibilities of an innovation, but it is difficult to foresee all its social consequences. The effects take shape through uses, organizations, institutions, and people's choices.

Let's think about the Internet. Born in a context of research and scientific communication, it has transformed commerce, information, relationships, and political participation. No one could have fully anticipated the platform society, just as today we cannot predict all the transformations that will accompany the evolution of artificial intelligence.

This does not mean giving up on imagining the future. It means recognizing that the future is not an inevitable consequence of technology: it is also the result of the decisions we make.

Artificial intelligences are many

Public debate today tends to identify artificial intelligence with large language models and generative applications. This is understandable: the ability to converse with a machine, produce texts, images, and programs has made AI accessible to millions of people.

But the history of artificial intelligence is much more complex.

Alongside data-driven learning approaches, there are systems based on the explicit representation of knowledge, logical reasoning, rules, and relationships between concepts.

These are not necessarily alternative approaches. One of the most interesting perspectives concerns precisely their integration.

Generative systems can produce responses and interpret information with extraordinary flexibility. But when they enter work processes, healthcare, education, or public services, the ability to generate a plausible response is not enough.

Reliability, verifiability, and responsibility are required.

The integration of statistical learning, explicit knowledge, and control systems can help reduce certain errors, limit permitted actions, and make decision-making processes more understandable.

Of course, no architecture alone guarantees reliability. The results must be verified in real-world usage contexts.

This is where technological research meets social responsibility.

The progress of AI requires the ability to learn, knowledge to interpret, and criteria to verify.

From the intelligence of machines to the quality of interaction

The spread of generative artificial intelligence has introduced a change that I consider particularly significant: the value no longer lies solely in the tool's functionalities, but in the quality of the interaction we are able to build.

Two people using the same system can achieve profoundly different results.

The difference depends on the questions they ask, the knowledge they possess, the context they can provide, and, above all, the ability to critically evaluate the answers.

Artificial intelligence can expand our cognitive possibilities, but it does not automatically replace judgment.

On the contrary, precisely because it makes it easier to produce information, texts, and solutions, it makes it even more important to know how to distinguish what is plausible from what is well-founded, what is useful from what is appropriate.

This change directly concerns education.

For a long time, we have primarily associated learning with the acquisition of knowledge and skills. Today, we must add a fundamental dimension: learning to interact critically with systems capable of processing information and proposing solutions.

It means knowing how to formulate questions, compare interpretations, verify sources, justify decisions, and recognize the limits of one's own knowledge.

It is not enough to learn how to use artificial intelligence. We must learn to think and work together with it, without giving up our autonomy.

It is a cultural challenge even before being a technological one.

Reliability: from the 2021 AI Act to the EMA model

A topic on which I wish to draw attention concerns the reliability of artificial intelligence systems, especially when they are used in critical contexts.

Already in 2021, during the European debate that accompanied the proposal of the AI Act, I had advocated the importance of introducing regulated testing environments, the so-called regulatory sandbox, in which to develop and verify innovative systems before their deployment.

This approach is linked to a principle I consider fundamental: technological innovation must be accompanied by validation processes proportional to the risks and continuous monitoring of the results.

In this perspective, I find the reference to the medicine evaluation model adopted in Europe particularly significant, in which the EMA, European Medicines Agency, plays a central role.

A drug is not evaluated solely for its therapeutic potential. It must undergo rigorous procedures that examine its quality, safety, and efficacy. And the evaluation does not end with authorization: it continues through the monitoring of effects during use.

Of course, an artificial intelligence system is not a drug. But the methodological principle offers an important insight.

When we entrust an intelligent system with tasks that can impact health, work, education, or access to services, we cannot limit ourselves to verifying its performance in the laboratory.

We need to understand its behavior in real contexts, identify its limits, evaluate its effects, and establish clear responsibilities.

The AI Act, whose initial proposal dates back to April 21, 2021, introduced a European perspective based on risk classification and the need to ensure reliable systems. The regulation of regulatory sandboxes represents one of the tools through which innovation, experimentation, and the protection of individuals can be reconciled.

The comparison with the EMA experience reinforces this approach: experiment first, verify during, and continue to monitor the effects after the introduction of a technology.

It is an issue that will become even more important with the spread of AI systems capable not only of generating responses but also of taking initiatives and performing actions.

The more a system can operate autonomously, the more precise we must be about the objectives, limits, checks, and responsibilities.

Reliability, therefore, is not just a technical characteristic.

It is a collective responsibility involving researchers, businesses, institutions, and citizens.

Knowledge cannot replace experience

However, there is another dimension, less discussed, which I consider fundamental.

In his book The Non-Things, the philosopher Byung-Chul Han reflects on a society in which information tends to prevail over the concrete experience of the world.

It is a particularly relevant reflection in the era of generative artificial intelligence.

We can ask a system to describe a forest to us, tell us about the characteristics of the trees, analyze its biodiversity, and illustrate the climate changes that threaten it.

We can obtain extraordinary information.

But crossing that forest is a different experience.

It means perceiving light, hearing sounds, recognizing smells, sensing temperature, engaging with the materiality of the world.

The two forms of knowledge should not be opposed. They can mutually enrich each other.

The problem arises when we confuse the representation of the experience with the experience itself.

In an increasingly intelligent system-mediated society, we will need to intentionally cultivate attention, sensitivity, and the ability to relate.

I think, for example, of mentoring and dialogue between generations.

An AI system can help us organize knowledge, identify career paths, or develop skills. But the relationship between a mentor and a younger person includes dimensions that are difficult to reduce to an exchange of information: trust, listening, mutual recognition, the ability to interpret experiences.

Technology can support these relationships. It should not impoverish them.

The real challenge is to expand human capabilities without diminishing autonomy, emotional intelligence, and the quality of social bonds.

Towards 2076: What I Expect and What I Desire

Imagining the world fifty years from now means distinguishing between two questions.

The first concerns what I consider plausible.

The second concerns what I wish to contribute to achieving.

I expect that by 2076 artificial intelligence systems will be much more integrated into the environments in which we live.

They will not only be tools for conversation. They will be able to perceive the context, interpret situations, collaborate with other systems, and perform actions in the physical world.

Let's think about assistance for elderly people.

We can envision robotic systems capable of performing logistical tasks, reporting abnormal situations, supporting healthcare personnel, and contributing to improving service organization.

But the quality of care will not depend solely on the performance of the robots.

It will depend on their reliability, the training of the people who use them, the organization of the services, and the ability to preserve the human relationship.

Above all, it will depend on the access conditions.

Who will be able to benefit from these innovations? Will they be available only to those with greater financial resources? Who will govern their use? How will the benefits, errors, and unexpected consequences be evaluated?

These are questions we need to ask ourselves today, not in 2076.

A company capable of learning at every age

If I think instead about the future I desire, my focus shifts from machines to people.

I would like a society in which everyone can continue to learn, change their career path, develop new skills, and contribute to collective life at every age.

A company capable of enhancing young researchers, offering them opportunities for growth and collaboration, without losing the wealth of experience of previous generations.

A society in which gender, origin, economic conditions, age, or disability do not represent obstacles to access to knowledge and opportunities.

Because innovation does not necessarily coincide with inclusion.

A technology can be extraordinarily advanced and, at the same time, widen existing inequalities.

For this reason, we cannot limit ourselves to measuring progress through productivity, processing speed, or the ability to automate tasks.

We should also ask ourselves whether innovations improve the quality of life, expand people's autonomy, strengthen relationships, and make participation more accessible.

It is a perspective that requires considering intelligences in the plural.

Human intelligence, emotional intelligence, scientific intelligence, artificial intelligence, collective intelligence, and organizational intelligence are not identical realities, nor are they interchangeable.

But they can contribute, through different forms of collaboration, to the construction of knowledge and the solution of complex problems.

The future will not be determined solely by the intelligence of machines. It will also depend on our ability to connect these different forms of intelligence.

Living the future with more knowledge, attention, and care

For over forty years, I have been observing the evolution of digital technologies and their relationship with learning processes, organizations, and society.

I have seen tools, languages, technological paradigms, and innovation models change.

One belief, however, has progressively strengthened: technology creates possibilities; it is people and institutions that turn them into progress.

We must not choose between enthusiasm and fear.

We must build the cultural, educational, organizational, and institutional conditions that allow the new technological capabilities to be used with awareness.

This means investing in research, continuous training, responsible experimentation, and participation.

It also means recognizing that knowledge is not just the accumulation of information, but the ability to interpret the world, exercise judgment, and take responsibility.

I expect that by 2076 we will have artificial intelligence systems far more capable than those of today.

I want a company equally capable of making them reliable and of transforming their potential into shared knowledge, autonomy, and inclusion.

I would like that in 2076 we know how to inhabit the world with more knowledge, more attention, and more care.

This future begins with the choices we make today.