I strongly disagree with this. Learning by doing is valuable and always has been. But it works best when people have a good understanding of what occupations and skills will be in demand. If there’s one thing we can be sure of, it’s that the future of the labor market is highly uncertain. Advising young people to abandon higher education in favor of an early entry into the labor market is misguided at best.
Jeffrey Hinton, widely considered one of the pioneers of modern artificial intelligence, once compared progress in his field to navigating in “fog”: you can see what is immediately in front of you, but not what is coming next. Accordingly, a major challenge for educators is to prepare students to work effectively in a fog-like environment. The solution is not to train them for specific tasks that may soon become obsolete, but to make them as adaptive as possible. Trying to train people for a fixed set of tasks when those tasks are constantly changing is a losing strategy. We need experienced drivers who can navigate unfamiliar roads and overcome unexpected obstacles.
From this perspective, education, and especially higher education, is more important than ever. Since we do not know what specific skills will be in demand in the future, it is necessary to go back to basics. A liberal education emphasizes how to think rather than what to do. It teaches students to reason, read carefully, write clearly, and evaluate evidence. These skills will be much more in demand than narrow technical competencies.
This does not mean ignoring technology. On the contrary, students should learn how to work with AI. But the goal should be to make them critical users and knowledgeable judges of AI tools, not passive consumers. It’s still important to teach the basics of math, logic, and reasoning, expose students to foundational texts, and teach them how arguments are constructed and tested. These are the skills that keep people ahead of rapidly evolving technology.
This principle raises two practical questions: what should we teach and how should we teach it? The first question is complex and will inevitably generate debate. While there may be broad agreement on the importance of basic concepts, the details will change over time. Our experience with earlier technologies provides useful guidance. The advent of calculators and computers has not eliminated the need to teach arithmetic. Students still learn how calculations work, but labor-intensive manual calculations are now delegated to machines. Similarly, spelling and grammar remain important, but software has largely replaced the need for endless exercises.
AI requires similar adaptation in many areas. LLMs are now very good at tasks such as summarizing text or highlighting main ideas – longtime staples of education. The same is increasingly true for programming, quantitative problem solving, and even writing. While these activities should not disappear from the curriculum, the goal must change. Students need to understand the underlying concepts and logic, rather than mastering each step of execution.
Successful students will be those who can effectively use AI tools to accomplish clearly defined goals. It’s the same with good management: success depends on prioritizing, structuring problems, and using available resources wisely. These are conceptual skills, not narrow technical ones.
The second, pedagogical issue concerns how learning is reinforced and assessed. Understanding takes some practice, but AI makes it easier than ever for students to avoid doing work on their own. Even highly motivated students will sometimes be tempted to take the path of least resistance, especially under time pressure. This is why we need a major change in the grading system. Homework essays, problem sets and unsupervised exams are becoming less and less effective. They need to be replaced by face-to-face tests and exams, oral assessments and real-time tasks, whether on paper or on the blackboard.
Such changes have far-reaching implications. They require face-to-face, smaller classes and more direct interaction between students and instructors. In many ways, it will mean a return to old models of learning, abandoning some of the scale and standardization introduced by earlier technologies. It may even usher in a new golden age of liberal arts education.
But this model also raises serious concerns. It places great responsibility on educators, who must be willing to apply standards and make difficult decisions. Educational institutions must support them in doing so. At the same time, assessment based on personal interaction raises valid concerns about bias. Standardized exams have their flaws, but their bias is at least noticeable. Subjective assessment based on oral exams and personal interaction may be less transparent.
Perhaps the most serious problem concerns inequality. Learning in small classes with a high degree of personalization is expensive. Elite institutions can afford it, but large public universities will struggle. Just as distance learning exacerbated the education gap during the pandemic, the shift to AI-intensive face-to-face instruction could disadvantage those who rely most heavily on public education.
Some argue that AI itself will reduce the need for formal education by providing information and personalized guidance on demand. But that assumes users know what to ask and how to interpret the answers. The most motivated or gifted people may excel in such an environment, but they would do it anyway. Formal education matters most to the broad middle class.
If AI is to benefit society, we will need more investment in education, not less. AI will replace jobs, but it will also create new ones. Education should be one of the sectors that will expand. As AI becomes widely available, the quality of education will depend less on access than on expectations and enforcement. Smaller classrooms, more teachers, and more personal interaction are all expensive, but the productivity gains promised by AI make such investments both possible and worthwhile.
Pinelope Kujanu Goldberg,
former chief economist of the World Bank Group and editor-in-chief of the American Economic Review,
is a professor of economics at Yale University.
© Project Syndicate, 2026.
www.project-syndicate.org
