The most interesting discussions about AI don’t always begin at major tech events. Sometimes they start over a beer and end up questioning the future of humanity.
There are conversations that begin as a simple exchange of ideas among friends and end up turning into something more.
That happened to me a few days ago in Valencia, in a bar near the Mestalla stadium.
I was sitting with two friends.
One teaches philosophy, the other works as a public administration officer.
Three very different professional paths, three different ways of observing the world.
As often happens these days, the conversation eventually turned to artificial intelligence.
The question was simple:
Are we witnessing a technological revolution destined to change everything, or a bubble driven by excessive expectations?
My philosopher friend had few doubts.
While acknowledging the potential of the technology, he saw in the current enthusiasm some typical signs of past technological fads.
His main concern, however, was not economic.
It was cultural.
According to him, the more we delegate cognitive tasks to AI, the more we risk losing the habit of thinking independently.
Less critical thinking.
Less analytical ability.
Less mental exercise.
A technology created to enhance productivity could, paradoxically, make us more dependent.
My friend, the public administration officer, looked at the phenomenon from a more practical perspective.
In his daily experience, he already sees processes, documents, and repetitive tasks that could be largely automated.
For him, the real question is not whether AI will change work, but how quickly organizations and institutions will be able to adapt.
Between the philosopher’s caution and the civil servant’s pragmatism, the conversation was becoming increasingly interesting.
From the table next to us, a man who had been listening for a while joined in.
He had a beer in his hand and, smiling, said:
“Sorry to interrupt, but I’ve been listening for a while. The topic is really interesting.”
He told us he was a software engineer originally from Palma de Mallorca.
He had worked in the field for many years and, despite having lived through several phases of digital transformation — from the rise of the internet and the dot-com bubble to the arrival of the modern smartphone — he claimed he had never seen anything comparable to what is happening today with artificial intelligence.
“We are only at the beginning.”
That was the phrase he kept repeating.
From language models to AI agents
What I found interesting was not so much his optimism, but the way he supported it.
He did not deny the limitations of current language models (LLMs).
On the contrary.
He emphasized how LLMs are probabilistic systems, very different from the way we are used to imagining human reasoning.
But the point, according to him, was not the individual model.
It was what happens when multiple AI agents begin to collaborate with each other, exchanging information, verifying results, and coordinating tasks.
His thesis was that from the collaboration of many artificial intelligences, new capabilities could emerge — difficult to imagine when observing each system in isolation.
At that moment I realized that the point was no longer AI as a single tool, but as an ecosystem of interacting systems.
A perspective that completely changes the scale of the problem.
The real limit of artificial intelligence: energy
From there, the conversation shifted to a topic that rarely appears in headlines: energy.
Human beings are an extraordinarily efficient machine.
With limited resources, we are capable of producing innovation, creativity, and progress.
Artificial intelligence, on the other hand, requires enormous amounts of computational power, infrastructure, and energy consumption.
Behind every technological advancement lies a very concrete question:
Do we have the necessary resources to sustain it?
Quantum computers and computational capacity
We ended up talking about quantum computers, qubits, and the possible future evolution of computational capacity.
A qubit, unlike a traditional bit that can represent only 0 or 1, can exist in a superposition of states until it is measured.
A feature that, at least in theory, could allow us to tackle extremely complex problems with computational power that is currently unimaginable.
Topics still far from everyday life, but potentially part of the next phase of this technological transformation.
Why the future of AI is so difficult to interpret
At that point, I paused to look around.
A philosophy professor.
A public administration officer.
A software engineer from Palma de Mallorca, met by chance over a beer.
Three different perspectives on the same phenomenon.
And perhaps this is exactly what makes AI so difficult to interpret.
It is not just a technological issue.
It is an economic, social, cultural, energy-related, and even philosophical one.
I keep thinking that we are facing a revolution comparable to the arrival of the internet or the great industrial revolutions of the past.
The difference is that, while we are living through it, it is almost impossible to predict its final destination with precision.
We can sense the potential.
We can imagine some scenarios.
We can discuss opportunities and risks.
But no one really knows what the world will look like in ten years.
And perhaps it is precisely this uncertainty that makes this historical moment so fascinating.
Conclusion
One thing I do know, though.
Every time I travel to Spain, I come back home with a few new ideas.
And sometimes the most interesting conversations don’t happen at conferences or major tech events.
They happen over a beer, in an ordinary bar, among people simply trying to understand where the next chapter of innovation will take us.
