In recent years, using artificial intelligence (AI) for virtually anything — from everyday tasks to complex scientific problems — has become a widespread trend. It is difficult to dispute that the emergence of AI represents a new kind of “industrial revolution”, one that affects not physical labour but intellectual work. Its presence and development are inevitable and unavoidable; however, its proper use is crucial.

Contemporary primary and secondary education increasingly prioritizes algorithmic, step-by-step thinking. As a result, higher education faces the complex task of teaching students — who have grown accustomed to convenience — to think, analyse, and filter information independently.

The university teacher of the future is effective only if they use AI in their teaching and teach students how to apply it responsibly and maximise its capabilities. At the same time, it must not be forgotten that current AI systems still exhibit a relatively high error rate and perform poorly on tasks requiring creativity. In higher-level mathematics, and more specifically in geometry education, AI is best described as a combination of a “super-Wikipedia” and a “conveniently accessible Wolfram Alpha”. This is precisely the point at which human contribution becomes indispensable: creativity, complex reasoning, and — above all — the evaluation of whether an obtained result is correct remain “human privileges”.

The current limitations of AI are well illustrated by a course offered to architecture students at the Budapest University of Technology and Economics: Advanced Computational Geometry. The course has a dual aim. First, students must determine, based solely on a few photographs (or occasionally video footage), which geometric surface types an existing building can be derived from – such as quadrics, surfaces of revolution, developable surfaces, or translation surfaces. This requires strong skills in form recognition and geometric analysis. Second, students must visualise their own design ideas without relying on classical modelling software (such as Rhino and Grasshopper). Instead, they use GeoGebra, a freely available software that is remarkably suitable for this purpose. Through this process, their knowledge of differential geometry develops significantly, which later enhances their efficiency when using professional modelling software.

Students are permitted to use AI when solving course assignments. The combination of geometric analysis and differential geometric modelling leads them to evaluate AI-generated outputs with a natural sense of scepticism: they immediately analyse, question, and correct the results. It is striking that in many cases students prefer traditional, manual methods of computation and routinely verify every result produced by software or AI tools. I believe the course has thus achieved its goal: we are training architects who possess excellent computational skills, who can master artificial intelligence rather than rely on it, and who retain all the — currently still — human privileges: critical thinking and creativity.

Today’s university students are the future engineers and scientists — those who may one day contribute to the development of AI itself. For this reason, maintaining independence from AI, and being able to rise above it when necessary, is a crucial component of an engineer’s intellectual toolkit.