A few days ago, a client asked me a simple question:
“In your view, what is the future of artificial intelligence in relation to humans?”
The question stuck with me.
I work in technology and use artificial intelligence regularly myself. I see what it makes possible today, but above all how fast its capabilities are evolving.
I don't believe artificial intelligence is fundamentally good or bad. I do believe, however, that we are building an extremely powerful tool, probably faster than our collective ability to decide how we want to use it.
And that is where my concern begins.
The first upheaval will be work
We often talk about AI as a technology of the future. Yet its impact on work is already a very present-day question.
Programming, graphic design, translation, writing, data analysis, customer service, research, marketing: a huge number of tasks that once took hours of work can now be sped up or partially automated.
That doesn't necessarily mean all these jobs will disappear. It's an important distinction.
The International Labour Organization estimates that about one in four workers worldwide holds a job with some degree of exposure to generative AI. But its conclusion matters just as much: in most cases, jobs are currently more likely to be transformed than to disappear entirely.
In my view, though, the change could be abrupt in some digital fields.
Why pay for several hours of work on certain tasks when one person assisted by AI can do them much faster?
And that question inevitably leads to another: what happens to the job market when one person's productivity is multiplied?
Physical jobs aren't necessarily safe
I used to think the divide was fairly simple: AI threatens digital jobs, while manual trades are relatively protected.
I'm no longer so sure.
The difference is probably more a matter of speed.
An AI can generate text or code instantly because it works in a digital environment. Replacing a physical task requires sensors, motors, energy, mechanical systems and a real-world environment that is far less predictable.
But robotics is advancing too.
Cleaning a floor, moving goods or performing certain repetitive tasks can gradually be automated.
On the other hand, a plumber facing a 40-year-old installation that matches no drawing, or a surgeon dealing with a rare complication, works in a far more complex environment.
Human experience, adaptability, judgment and knowledge built up over years remain extremely important.
So we should probably stop asking “which jobs will AI replace?” and ask instead “which tasks within this job can be automated?”
The paradox: we can hardly choose not to use it
Then there is a much deeper problem.
Imagine two employees whose boss asks them the same question.
The first looks for the information in several books, goes through various documents, compares them and comes back with an answer a few hours later.
The second uses an AI, checks the relevant sources and comes back with a summary in a few minutes.
Even if the first refuses AI on principle, how long can they really hold that position if their colleagues become much faster?
It's a bit like what happened with the Internet.
We never collectively decided that paper encyclopedias were bad. We simply found an infinitely faster way to access information.
AI pushes that logic even further: it no longer just gives us access to information, it can help us analyze, summarize, transform and use it.
Not using AI could therefore eventually become a professional disadvantage.
What really worries me: the race
My main concern, though, isn't ChatGPT, Claude or any other particular tool.
It's the competition.
The companies building the large models are competing. The companies using those models are competing. And countries themselves are competing.
So everyone has an excellent reason to keep going.
If one company voluntarily decides to slow down, another can carry on. If one country imposes significant restrictions, another can choose a different approach.
That is what makes genuine global coordination so difficult.
And the concerns no longer come only from science-fiction scenarios. In September 2026, the resignation of Anthropic researcher Jacob Coxon publicly reignited the debate about the risks of highly advanced AI systems and the speed of their development. The catastrophic scenarios he describes remain contested and are obviously not established predictions.
Still, it raises an interesting question:
If even some of the people working directly on these technologies are calling for more thought about their limits, shouldn't we at least be having that conversation collectively?
AI is like a knife
That's probably the image that best sums up how I see artificial intelligence.
A knife can be used to prepare a meal and feed a family. The same knife can be used to hurt someone.
The knife has no intention.
It's a tool.
Artificial intelligence is obviously infinitely more complex, but the principle holds: the same technology can create enormous value and new risks at the same time.
- It can help a doctor.
- It can help a programmer.
- It can speed up a scientific discovery.
- It can give knowledge to someone who never had access to it before.
But it can also be used to manipulate, deceive, automate malicious activities or simply make bad decisions at a previously impossible scale.
So the problem isn't only the tool. The problem can be whoever is using it.
And history shows that humans don't always use powerful tools wisely.
Humans still have something AI doesn't
There is, however, one dimension I find important in this discussion.
AI learns from information produced by humanity. It can combine knowledge, recognize patterns and produce something new from what it has learned.
But it hasn't lived my life.
It never had my first job, made my mistakes, met my clients or spent twenty years accumulating my experiences.
This is especially clear with experts.
A specialized surgeon may have faced highly unusual situations that were never documented on the Internet. Their expertise isn't just information: it is also made of thousands of hours of experience.
The same is true of creative work.
A person tells a story differently because they have lived through something. They have memories, a culture, relationships and a personal context.
AI can reproduce a great deal.
But reproducing information and having lived an experience are not the same thing.
So, is AI a danger to humans?
I think the wrong question would simply be: “Is AI dangerous?”
The question that interests me more is:
What will humans do with a technology whose capabilities are growing extremely fast?
Because I don't believe we will stop AI.
Economic competition is too strong. Competition between countries is too strong. And the productivity gains are simply too attractive.
Nor do I want us to abandon this technology. Its potential benefits are extraordinary.
But between stopping everything and moving ahead without limits, there is a huge space for reflection.
We should be talking more about safety, responsibility, education, transparency and the limits we collectively want to set.
Because in the end, the question may not be how far artificial intelligence can go.
The real question is how far we, as humans, want to let it go.