Tech • AI • Robotics • Game
Does AI Make Us Smarter or More Stupid?
5/10TL;DR
Artificial intelligence is less likely to make everyone uniformly smarter or dumber than to widen existing cognitive and social gaps, rewarding people who already know how to question, verify, learn and communicate.
KEY POINTS
The debate misses the real issue
Public argument often swings between two simplistic camps: AI as revolutionary genius or AI as useless error-prone hype. That framing obscures the practical reality that outcomes depend on use, context, profession, and existing human skills. Since ChatGPT reached the public in November 2022, systems have improved rapidly across language, code, vision, voice and math, while limits and risks have remained real.
The Matthew effect offers a useful lens
The argument draws on the Matthew effect, a concept formalized by sociologist Robert K. Merton in 1968 and later applied to education in 1986. The principle is simple: those who already possess an advantage tend to gain more, while those with fewer resources fall further behind. In reading, for example, strong readers build vocabulary and comprehension faster, while weak readers lose ground; the same dynamic may now apply to the use of AI.
The divide may be cognitive more than financial
The main fracture is not presented as a simple battle between rich and poor or between people who can and cannot afford access. Powerful models are already widely available, sometimes for around $20 a month, and some governments have considered free access for all. The deeper divide lies in whether users can formulate good questions, challenge answers, seek evidence and turn conversation into learning.
AI can function as either tutor or echo chamber
For a curious and disciplined user, an assistant in a phone can act like a teacher, coach, explainer or research partner. For a passive or dogmatic user, the same tool can become a flattering accomplice that reinforces bias, poor reasoning and false certainty. In that sense, AI can amplify both intellectual rigor and intellectual laziness.
School and civic culture are poorly prepared
The skills that matter most for productive AI use are not only technical. They include listening, debating, self-questioning, choosing precise words, asking follow-up questions and tolerating uncertainty. These habits are described as insufficiently taught in schools and weakly modeled in a media environment shaped by speed, polarization and slogans rather than careful inquiry.
Task disruption is spreading beyond early targets
The impact of AI is described less as whole professions disappearing overnight than as tasks being replaced piece by piece. Creative work, assistance functions and support services are already under pressure, while many workers still assume their own role is safe. The warning is that teachers, executives and politicians may be underestimating how fast capable autonomous systems can take over analysis, coordination and decision support.
Math advances have intensified the shock
Recent claims from OpenAI are cited as emblematic of the new level of disruption. Among them are reports of progress on a Millennium Prize problem and the release of 722 handwritten-style proofs of published mathematical results. Whether every claim holds up to scrutiny or not, the broader effect has been to unsettle parts of the scientific world by showing that prestigious cognitive work is no longer obviously protected.
Cédric Villani reflects the wider reckoning
Cédric Villani, the 2010 Fields Medal winner and author of a major 2018 French report on AI, is presented as a case of intellectual reappraisal. After long skepticism about whether systems could truly reason, he has publicly acknowledged being shaken by recent progress. His warning is stark: for a motivated student, AI offers personalized support, but for an unmotivated one it can weaken effort and deepen dependence.
The policy choice is between restriction and reinvention
One response is defensive regulation aimed at slowing deployment, often justified by fears over manipulation, democratic harm and the environmental cost of data centers. The alternative presented is a broader civic project: democratize access, teach the skills needed to use AI well, and launch a new humanistic push comparable to an Age of Enlightenment. The central claim is that limiting the tools without upgrading people would leave the underlying weakness untouched.
CONCLUSION
The strongest case is not that AI automatically makes people stupid, but that it magnifies what users already bring to it. The urgent challenge is therefore educational and political: build the human capacities that let societies use AI as a lever for judgment rather than a machine for deeper division.
Ask a question