As generative AI reshapes how we learn, universities should encourage academic freedom, critical inquiry, and international cooperation while helping society navigate the opportunities and risks of technological change.
While we are living in a time of accelerated change, a changing world in itself is nothing new, given that humans have always been innovative and progressive. For example, the invention and introduction of the typewriter took office automation to a new level, drastically changing and accelerating the way humans worked. Similar developments occurred with the introduction of the personal computer and the invention of the World Wide Web, which ushered in the platform industry. All these transformations increased work efficiency, accompanied by a substantial shift in competencies and knowledge required by a modern workforce. What renders the current transformation unique is its pace, magnitude and scale. In an increasingly interconnected world, change is happening at a rate never seen before, impacting lives directly and indirectly. This is especially true in the case of Generative Artificial Intelligence (GenAI), where the technology is progressing faster than any regulatory or administrative framework can keep abreast of.
University of Zurich, Switzerland.
The Sovereignty of Interpretation
The current debate about GenAI is marked by contrasts – the extreme expectation of the new technology’s potential, juxtaposed against the extreme dangers, and risks it presents. As is so often the case, the answer lies in neither extreme. Rather, it is a matter of interpretation. This is where universities must take a stand and maintain the sovereignty of interpretation through quality research, teaching, and services. Accelerated change requires interdisciplinary and global collaboration, which requires a collective of thinking cultures and diverse methodological approaches applied to integrate new technologies into our daily lives.
GenAI is certainly one of many current global challenges and therefore can only be addressed globally. In general, science has always benefited from global collaboration. This is also the case in the field of GenAI. International cooperation, however, cannot be understood as openness without conditions. Sustainable academic partnerships depend on academic freedom, institutional autonomy, reciprocity, and research integrity. These principles are not barriers to cooperation; they are what make meaningful cooperation possible. Universities do not just educate future leaders and thinkers. They also provide spaces for open discourse, for reflection and more importantly for ambiguity and uncertainty. This is especially relevant in the current times of accelerated change, where universities can be a space to reflect and think critically. They should preserve channels for academic dialogue wherever responsible cooperation is possible, while safeguarding academic freedom and institutional autonomy.
Should Universities Be Neutral?
While the concept of neutrality is often associated with Switzerland, the neutrality of a university should be understood in a distinct sense. It is not about political neutrality, but about preserving the institutional conditions that allow scholarship, debate, and self-reflection to flourish. Therefore, neutrality must not be mistaken for indifference or ignorance. As stated in the University of Chicago’s 1967 Kalven Committee’s report on the University’s Role in Political and Social Action: “The neutrality of the university as an institution arises then not from a lack of courage nor out of indifference and insensitivity. It arises out of respect for free inquiry and the obligation to cherish a diversity of viewpoints.” Last year, I elaborated on this concept in more detail in my opinion essay titled “Universities Should be Upsetting – Let’s Resist the Calls for Conformity.”
The laboratory building on the Irchel campus of the University of Zurich. The UZI 5 building combines sustainable and energy-efficient architecture with state-of-the-art research infrastructure.
Contingencies and Opportunities
It is instructive to remember that GenAI in its current form is contingent: There could have been and there still are alternatives. This may seem to contradict the frequently expressed view that “AI is inevitable.” The development of AI has seen waves of progress, with valleys of stagnation in between. It has raised hope and simultaneously disappointed many. Large Language Models (LLMs), which are the basis of GenAI, are built on big training data sets and predict answers via best guesses and probabilities. They are especially suitable for tasks where knowledge gaps can be filled or complemented with GenAI’s strengths, such as translation and summarization. However, they don’t have an intrinsic kernel of truth, making it inherently challenging to trust if they are factually and morally correct. Physical and causal reasoning will remain a challenge for these models. Or as Yann LeCun, Turing Award winner and one of the “godfathers” of AI has stated in an interview: “The digital world and the world of discrete things is relatively easy to manipulate for current technologies. But dealing with the real world is much harder.”
Hence, LLMs have yet to earn their first “L,” which stands for “large.” There is still a long way to go, in particular properly representing lexical and grammatical diversity in a wider context, as well as using the correct vocabulary and grammar in the correct context. But that is not enough. Basic, formulaic wordplay or simple jokes can be generated by GenAI through matching phonetic patterns and semantic context. But a genuine cognitive response to double meaning, let alone understanding of a subtle play on words, still lies ahead. A sufficient plurality of viewpoints and styles will improve the answers of GenAI, requiring global collaboration of researchers and universities. To maximize the diversity of input to the models, universities must increasingly serve as places of exchange for people from different cultural and scientific backgrounds.
Friction and Ambiguity
For humans to learn well, two things are needed: A good balance between competence and knowledge, as well as some sort of friction, triggering independent thinking. The current technological change has induced a subtle shift from knowledge-based learning to competence-based learning. Past technical revolutions have shown a recurring pattern: After the apex of the transformation process, people have acquired the competencies to become proficient in the new technology, leading to increased efficiency. In parallel, a reverse shift to knowledge-based learning can be observed – but on a completely different knowledge base to that previously seen. Currently, we are impressed how helpful GenAI can be in making people acquire factual knowledge faster. Consequently, it is less about knowledge-based learning and more about the competence of using GenAI well.
However, this new way of working and studying can also be very detrimental to the learning process. In general, learning happens at a place of “constructive friction”: When we are faced with contradictions and doubts, we must think for ourselves and learn new things. While GenAI is very convenient in acquiring new knowledge, it eliminates possible points of friction. Therefore, it is the universities’ task to defend these frictions. With GenAI changing the way we learn, we must be ready to anticipate changes in the learning processes by keeping the sovereignty of interpretation. As this process is not achieved in isolation, it is increasingly more important for universities to commit to collaboration and interdisciplinarity, and to increase humans’ tolerance of ambiguity.
MICHAEL SCHAEPMAN is president of the University of Zurich, Switzerland.
