Artificial intelligence to think with us, not to think for us

We have access today to something extraordinary. In a matter of seconds, an artificial intelligence can explain a complex concept to us, summarize hundreds of pages, test ideas against one another, translate a text, analyze data, propose hypotheses, or guide us as we learn about a subject we knew almost nothing about a few hours earlier. Never before has so much knowledge and intellectual capability been so readily available to so many people.

That power, however, confronts us with a question that is far more personal than it appears. We can use artificial intelligence to develop our own thinking, explore further, learn faster, and challenge our ideas. We can also fall into the habit of transferring to it a growing share of the intellectual effort we would otherwise have made ourselves. From the outside, both uses can produce a comparable result, while their effects on our understanding and our skills are profoundly different.

In an interview with 98.5, neurosurgeon and neuro-oncologist David Fortin expressed precisely these concerns about the possible consequences of excessive use of artificial intelligence on our cognitive abilities and our critical thinking. His reflection echoes a question that also appears in various studies devoted to AI and learning: how do we benefit from a tool capable of assisting us intellectually without gradually surrendering to it the very exercise that develops our intelligence?

This question deserves to be approached with confidence rather than fear. We have a technology capable of considerably amplifying some of our intellectual abilities, and we can choose the relationship we want to build with it. The challenge is to learn to think with artificial intelligence while continuing to develop our own capacity to think.

Our brain develops by working. Searching for a word, attempting to solve a problem, mentally organizing an idea, or trying to understand a concept sometimes requires a few minutes of discomfort. We do not find the answer immediately, we revisit our reasoning, make connections, make mistakes, and try a new approach. That intellectual friction is part of the process through which we build our knowledge and develop our ability to reason.

Artificial intelligence can now remove a large part of that friction. An answer that would have required research, reflection, or several attempts can be obtained almost instantly. That ease is one of the technology’s great strengths, particularly when it allows us to get past a pointless obstacle or to quickly access knowledge that would otherwise have been difficult to find.

It becomes more delicate, however, when we systematically transfer to it the cognitive effort that would have contributed to our learning. The OECD is paying particular attention to the phenomena of cognitive offloading and metacognitive disengagement that can appear when general-purpose AI tools take on too large a share of intellectual work. The same technologies can nonetheless produce far more interesting results when they are used with the explicit intention of supporting learning.

The relevant question is therefore not simply whether we use artificial intelligence a lot or a little. It is rather to understand what we ask it to accomplish within our thinking process.

Take a complex subject we want to understand. We can immediately ask an artificial intelligence to produce a complete analysis, then adopt that analysis because it appears convincing and well structured. The result can be excellent and save us a great deal of time, while our personal understanding of the problem will have advanced relatively little.

We can use exactly the same tool in a very different way. We can begin by thinking through the problem, articulating what we understand of it, identifying our uncertainties, and building an initial position. Artificial intelligence can then challenge our reasoning, look for what we have overlooked, present the arguments that contradict it, identify the knowledge we lack, and suggest additional sources or avenues. We can then take up our analysis again with more information and a deeper understanding. In both situations, the artificial intelligence did the work. In the second, it also made us work.

This distinction appears in research devoted to knowledge workers. A study published in 2025 by researchers from Microsoft Research and Carnegie Mellon examined hundreds of examples of AI use reported by workers. The researchers observed in particular that high confidence in artificial intelligence was associated with less critical thinking effort, whereas greater confidence on the part of individuals in their own ability to complete the task was associated with more critical reflection. They also found that AI shifts part of the intellectual effort toward verifying, integrating, and supervising what it produces.

That shift can represent a tremendous evolution in our capabilities. We do not necessarily need to carry out each step of a task manually in order to keep exercising our intelligence. We do, however, need to retain enough knowledge, understanding, and judgment to supervise what has been delegated. The more the machine takes over execution, the more important our ability to understand the result and assess its quality becomes.

This perspective makes it possible to imagine a far more stimulating relationship with artificial intelligence. Instead of using it primarily as an answer machine, we can make it an intellectual training partner. It can explain a concept and then ask us questions to check our understanding, look for the flaws in our reasoning, present the best argument against our position, provide a hint rather than the complete solution, or compare our analysis with several other ways of approaching the same problem.

We can also ask it to temporarily defend a position we disagree with in order to test the soundness of our convictions. A conversation with AI then becomes an opportunity to encounter objections we had not considered and to discover the places where our own reasoning deserves to be developed further.

This use aligns with the principle of human agency put forward by UNESCO in its work on artificial intelligence. A technology capable of generating ideas and explanations can serve to challenge and extend human thinking when its use preserves the user’s ability to understand, choose, and exercise judgment.

Artificial intelligence also has extraordinary potential as a learning accelerator. A curious person can today begin exploring quantum physics, economics, programming, history, philosophy, a foreign language, or practically any field and receive explanations adapted to their level of knowledge. When one explanation does not work, they can ask for another, request an analogy, get an example, try an exercise, and then ask why their answer was incorrect.

An expert can use the same capability in the opposite direction and explore the frontiers of their own field more quickly. They can test hypotheses, bring disciplines together, seek out different perspectives, or use AI to rapidly survey an intellectual territory before focusing their expertise on the most relevant elements.

We thus have unprecedented access to a form of personalized intellectual support. Used with curiosity, artificial intelligence can become a remarkable learning machine: one answer leads to a new question, which leads to an unfamiliar concept, then to a theory, a person, a discipline, or a new line of inquiry. A few hours later, we can understand something we were unable to explain before.

That capability, however, makes preserving our own knowledge even more important. The more convincing artificial intelligences become, the more we need enough understanding to recognize an inconsistency, question a claim, and detect the limits of an answer. A confident, elegant formulation is never a guarantee of truth.

We must in particular avoid a circular dependency in which we would gradually stop learning because artificial intelligence knows the answers, and then become unable to evaluate those same answers because we have stopped acquiring the knowledge needed to understand them.

Human judgment does not operate in a vacuum. It is built from knowledge, experience, comparisons, past mistakes, an understanding of context, and exchanges with others. Our general culture therefore retains all of its value. Reading remains precious, writing for ourselves remains precious, and learning a language, understanding history, developing professional expertise, arguing with other people, and sometimes solving a problem without assistance remain essential exercises.

Artificial intelligence adds an extraordinary capability to that set. It can increase the speed at which we learn and the breadth of the knowledge we can access. These new possibilities take on their full value, however, when they rest on human foundations solid enough to allow understanding and judgment.

This reflection also has a more intimate dimension. Our thoughts, our opinions, and our way of understanding the world help define who we are. They are built from our history, our experiences, our reading, the people we meet, our values, our mistakes, and the discussions we have with others. We sometimes change our minds because we discover new information or because someone compels us to look at a question from an angle we had never considered.

Artificial intelligence can now take part in that process. This possibility is fascinating, since it gives us access to a considerable diversity of perspectives and can help us step out of certain blind spots. At the same time, it requires that we remain aware of the difference between receiving an additional intellectual influence and gradually transferring to a machine the formation of our own judgment.

An AI can present us with ten different ways of interpreting a situation, identify the strengths and weaknesses of each, and show us arguments we would never have thought of. That richness can considerably improve our thinking. The decision about what we consider right, desirable, or consistent with our values nevertheless remains profoundly human.

This responsibility takes on particular importance as systems become more personalized. An artificial intelligence able to know our context, our habits, our knowledge, and our preferences can become an extremely powerful intellectual partner. That closeness makes it even more necessary to be able to recognize when we are using the tool to broaden our thinking and when we are simply beginning to accept its conclusions.

Organizations will face this question as well. As employees work with copilots and agents, productivity should not be the only indicator being watched. It will also be necessary to ask which human skills must be preserved, which knowledge must continue to be mastered, and in which situations a person must remain able to perform, or sufficiently understand, the work done by the machine.

AI Adoption and organizational transformation will therefore have to take this dimension into account. Augmenting a team with artificial intelligence also means thinking about how responsibilities, knowledge, and judgment evolve. The goal is to enable people to exercise their expertise at a higher level thanks to the capabilities of machines while preserving the understanding needed to supervise them and to make the decisions that remain their responsibility.

This approach corresponds to a profoundly human conception of artificial intelligence. The challenge is not solely to determine which tasks can be automated. It is to identify the best possible division between what machines can accelerate and what humans must continue to learn, understand, question, and decide.

Artificial intelligence is becoming more capable, and that evolution offers us extraordinary intellectual possibilities. We can access knowledge more quickly, explore more ideas, test our reasoning, learn new disciplines, and have almost continuous access to a partner capable of helping us progress.

How we choose to use that power will nevertheless make a considerable difference. When it becomes a permanent shortcut to the answer, AI can spare us certain intellectual efforts that are precisely what develop our skills. When it becomes a teacher, a challenger, an explorer, a thinking partner, and an amplifier of knowledge, it can on the contrary give us access to learning opportunities that no previous generation has had on this scale.

Maturity will probably consist in knowing how to recognize the difference. Some tasks can be handed over almost entirely to a machine because performing them manually adds little value. Others deserve that we maintain active intellectual involvement because they build our knowledge, our expertise, or our judgment. The challenge will be learning to consciously choose what we wish to delegate.

At Quantum Beyond, our vision of human-centered artificial intelligence necessarily includes this dimension. Governance, cybersecurity, data protection, AI Onboarding, and control mechanisms are essential for framing systems. Human and organizational transformation is just as essential, because a truly useful AI should augment the capabilities of the people who work with it and allow them to exercise their expertise more fully.

We now have machines capable of searching faster than we can, processing considerable volumes of information, exploring thousands of possibilities, and supporting our thinking almost instantly. We have every reason to take full advantage of these capabilities to learn more, understand more deeply, test our ideas, and push back the limits of what we know.

Our thinking nevertheless remains a territory we have every interest in continuing to cultivate. While artificial intelligence can help us explore much farther, the understanding, judgment, and responsibility we develop along the way retain a profoundly human value. The true potential of this technology will emerge when its power allows us not only to obtain better answers, but also to become better at asking questions, understanding what they mean, and deciding for ourselves what we want to do with them.