Abstract about GenAI and Higher Education by Professor Jeanette Hofmann

Unlike many other innovations, the development of genAI affects the core tasks of higher education in a comprehensive and cross-disciplinary manner. In the summer of 2025, OpenAI compared the capabilities of its latest language model, GPT-5, to a conversation with “PhD level friends” — an intellectual achievement that university graduates typically don’t reach. Does this imply that universities are facing an automation and thus a devaluation of their educational content?

It seems important to underscore the profound uncertainty currently surrounding AI. The assessments of machine learning technology diverge widely, even within academia. While critical AI studies increasingly emphasise the intellectual limits of machine learning and the growing hype about AI, other disciplines celebrate the imminent performance growth towards an “Artificial General Intelligence” accompanied by enormous advances in productivity.

One of the central limitations of genAI lies in the probabilistic nature of its output. What has been referred to as the “stochastic parrot” (Bender et al. 2021) means that current approaches of machine learning systems lack any concept of what higher education excels at, comprehending meaning, questioning assumptions, and thereby pursuing new insights or even truth. With higher education and genAI, two opposing paradigms of knowledge generation and quality assessment confront each other and now must endure continuous comparisons. At stake here is, not least, the belief in the non-substitutability of human knowledge production. In view of the increasing political attacks on the independence of higher education in many countries, a crucial task will be to maintain institutional trust in the quality standards of academic knowledge generation.