People Come First

People collaborating around laptops in a workspace.

A group of people sit around a wooden table working on laptops in a warm collaborative workspace, symbolizing human-centered technology, teamwork, and the idea that artificial intelligence should serve people rather than replace them.

You don’t need to be a technician to use technology. What we need is a humanistic approach — one that keeps people and what people actually do at the center of everything.
— Corrado Giustozzi
Fernando De Haro People Come First

Fernando De Haro - People Come First. AI Requires a Humanistic Culture.

Corrado Giustozzi is one of the experts who participated in the presentation of Pope Leo XIV’s first encyclical at the Vatican. A strategic consultant, public speaker, and educator on cybersecurity and information security, he has also collaborated with the European Union.

Giustozzi argues that, contrary to popular belief, AI does not operate algorithmically but probabilistically. He contends that the sweeping corporate consolidation now reshaping the technology sector is not a consequence of AI itself. The power wielded by Big Tech — surpassing that of many nation-states — is rooted in other factors entirely.

How do you assess the encyclical Magnifica Humanitas?

It’s hard to do it justice in just a few words. First and foremost, I think it’s enormously significant that this encyclical was written right now, on these issues — and the repeated references to Rerum Novarum strike me as very apt. Times are changing; new technologies are reshaping everyone’s lives and society at large, and I find it genuinely important that the Church is stepping up to address this.

Christopher Olah, an AI pioneer and Anthropic co-founder, said when he presented the encyclical that we don’t fully understand how it works. The encyclical itself acknowledges this. For a layperson, that’s a bit mind-bending.

And that, in my view, is the central stumbling block in any attempt to understand the issue. There’s an apparent paradox: how can a system built on algorithms — which by definition should be deterministic, rule-bound, and transparent — exhibit behavior that seems incomprehensible? It appears, in other words, to behave non-deterministically. This throws people off.

The misunderstanding that most laypeople never catch is this: the behavior of an AI system is not determined by its algorithms. Standard computing — any conventional automated process — works by encoding rules to solve a specific, well-defined problem. If I need to calculate a square root, I know the procedure: I encode it as an algorithm, and that algorithm will always return the correct answer. If I need to calculate someone’s monthly paycheck, I know the rules: days worked, base salary, bonuses, deductions — a whole set of factors — and I apply straightforward formulas to get a precise result.

AI doesn’t work that way. The solution to a problem is not encoded in an algorithm, because some problems simply don’t have algorithmic solutions. Take this one: how do you tell a dog from a cat? What’s the rule? I’ve just stumped you. There is no clean, codifiable rule for distinguishing a dog from a cat. You might say, “They have four legs” — both do. “They have whiskers” — both do. “They come in many colors” — both do. In other words, there is no precise, algorithmic pathway to telling them apart.

So if I want to build a machine that can recognize dogs from cats, I show it images. I can’t encode the solution as a rule. But what I can do is encode a different problem: how does a system learn from experience? I build a learning system and feed it thousands of photos of dogs and cats — every color, every size, every angle, every backdrop — and each time I tell it: “This is a dog. This is a cat.” Eventually, after processing enough examples, the machine begins to abstract the essence of “dogness” and “catness.”

(Forgive the neologisms — I’m not sure what else to call it.) It has learned that all those images of a certain type correspond to dogs, and the others to cats. At that point, I show it a new photo and ask, “What is this?” And it tells me, statistically: “In my best judgment, that’s a dog” or “In my best judgment, that’s a cat.” If the training was solid, it gets it right. But this is probabilistic — not algorithmic. This is how AI works. And when you ask it, “But why did you say it’s a dog?” the machine has no answer.

The Pope says AI is not neutral and also flags the fact that its development is being driven by large corporations that are accumulating enormous power. Are the two things connected?

Personally, I still believe technology is neutral — it’s how we use it that makes the difference. So it’s not that AI, in and of itself, drives the concentration of power. Had another disruptive technology come along first, the same thing would have happened. The consolidation of power is independent of the technology. Cars are a perfect example: enormous power concentrated in oil companies and manufacturers — not because of the car itself, but because of how the industry evolved around it.

In this case, it’s not just a risk on the horizon — it’s already baked in. I was reading just this morning that Anthropic’s valuation has reached $500 billion — an absolutely staggering figure. This is, without question, a pressing social issue that demands serious public debate. High tech in general — AI is merely the cherry on top — has produced a concentration of power on a scale the world has never seen before.

How do we put a check on this? States are weaker than these companies.

Most of these companies are more powerful than entire nations — their market valuations exceed the GDP of many countries, and their reach is enormous. This is a genuinely serious problem, and I honestly don’t know what the answer is. The old antitrust playbook likely no longer applies. Nations no longer have the leverage to impose meaningful limits on companies that are, as you rightly point out, stronger than the nations themselves.

It seems to me this isn’t solely a matter of ethical guardrails and state sovereignty. What may matter even more is the kind of human being we need to cultivate in the face of AI’s development. What kind of person, what kind of education, do we need to foster in order to be free and make better use of AI?

I believe the answer is, first and foremost, a humanistic culture. I remain convinced that the humanistic approach is where we have to start.

Why?

You don’t need to be a technician to use technology. Take the car as an example: either you’re a mechanic — in which case you need to understand how the engine works — or you just need to know how to drive safely, get the kids to school, and head to the beach on a Sunday, without endangering yourself or anyone else. In that sense, the world has changed enormously. At the beginning of the last century, driving a car required genuine mechanical expertise, because cars broke down constantly and you had to fix them yourself. Today, it’s perfectly normal for a grandmother to drive her grandchildren to school in the morning — not only because the cars themselves have improved beyond recognition, but because an entire ecosystem has grown up around them: traffic laws, road signs, better roads, driver’s education. Society has built a whole framework of standards, rules, and systems that make using the technology accessible to everyone — no expertise required, unless you’re a mechanic or an automaker by trade.

The same arc played out in computing. In the 1970s and ’80s, it was widely said that anyone who didn’t know how to program a computer would be the illiterate of the year 2000. That turned out to be wrong. You don’t need to be a computer scientist to use a computer or a smartphone.

What we need instead, in my view, is a humanistic approach. These technologies are tools that support us in our daily lives. Once again, if you’re an engineer, you use a computer to calculate load-bearing structures; if you’re an ordinary person, you use it to manage your calendar, look things up, or organize your files. It’s as if you had a brilliant human assistant working alongside you to help solve your problems. People — and what people actually do — are at the center of everything. An approach that is too machine-centric is simply inefficient, whatever the philosophical framing. It’s like someone who listens to music on a high-end audio system but isn’t actually listening to the music: they’re listening to the equipment. They’re fixating on whether the system sounds right and missing the music entirely. A great sound system exists so you can hear the music. The point is to listen.

The question of a humanistic education gets at the meaning of what one does and what one knows. But here’s the problem: AI gives the appearance of meaning things. That’s exactly where the ambiguity comes from.

And that’s precisely why a humanistic approach matters — to put everything back in proper perspective. We absolutely need to explain the limits of these technologies. Right now, amid all the hype and sky-high expectations, people assume AI can do anything, that it’s flawless. It’s not. No technology is. AI has structural limitations that can always be pushed further — but never fully eliminated. So we need to be able to tell users clearly: here’s where this tool works well, and here’s where it falls short.

In other fields, this is much more intuitive. Sticking with the automotive world: you can’t take a compact city car onto a racetrack. And if you’re a carpenter, you can’t haul your tools in a Ferrari. Every tool is built for certain tasks. For the record, I think it was a mistake to call it “intelligence,” because that’s precisely what it isn’t. It has impressive capacities for analysis and pattern recognition, but jumping from that to calling it intelligent is a stretch. The name alone generates expectations that can never be met.

But when you talk about limits, it seems to me you’re pointing to something deeper than ethical constraints — limits of conception, of meaning, of what technology actually is, and of what it means to be human. Based on what you’re describing, this isn’t really an ethics problem. It’s a question of understanding the human subject.

What I’m pointing to first are technical limits — things AI genuinely does poorly. The famous “hallucinations” are one example, but there are many others. At this stage, AI still often fails to draw correct inferences in certain situations. It is extraordinarily effective when asked to analyze documents — even very large and complex bodies of text — and to surface structures and relationships. It is far less effective at genuine invention, because at this stage it typically ends up producing something that riffs on what already exists, and that doesn’t mean it’s right. I’m speaking in broad strokes, of course — there are countless different applications.

That said, I wasn’t specifically talking about ethical limits just now, but those clearly exist as well. Using AI — or any advanced automation tool — uncritically can lead to discrimination, bias, and outright errors. In that sense, AI sharpens a problem that already existed: whenever we try to over-automate processes that involve decisions affecting real people’s lives, we risk generalizing in ways that discriminate or introduce bias. This is certainly important, and it’s an issue the European Union’s AI Act addresses head-on. When we talk about technologies being more or less “high-risk,” we are essentially talking about this: the capacity to infringe on people’s rights and freedoms through fully automated decision-making.

Fernando De Haro

Fernando de Haro is a Spanish journalist, academic, and radio director at COPE. With degrees in journalism, law, and a PhD in information science, he's known for documentaries on Christian persecution. De Haro explores religion's role in society through his media work and publications, including a book on Don Giussani's life.

Previous
Previous

No Hidden Agenda

Next
Next

Under the Spell of God