AI is forcing universities to rethink what a degree is for
As machines take over more knowledge-based tasks, higher education is being pushed toward critical thinking, practical skills, mentoring and new forms of assessment.
The race between universities and AI has already begun. On one side are tools that can write papers, generate code and provide answers within seconds. On the other are institutions that still rely heavily on lectures, assignments and knowledge-based exams.
Academia understands that these traditional methods are no longer enough, but the solution is far from simple. An oral exam can establish whether a student actually understands the material they submitted, but conducting one becomes extremely difficult in introductory courses with hundreds of students. AI is therefore emerging not merely as a technological challenge for universities, but as a fundamental test of how higher education operates.
This is not a crisis of demand for higher education. It is a crisis that requires universities to rethink each of their core components: the profile of the graduate, the learning process, the teaching process, methods of assessment and research.
One of the central problems is a mismatch within the academic workforce itself. Academics are promoted primarily on the strength of their research, while many view teaching as an obligation and are content to convey their knowledge through lectures. In the AI era, however, they will increasingly be expected not only to teach, but also to educate, guide and mentor.
There is another problem. If any written assignment can potentially be produced with the help of AI, oral evaluation may become one of the most important ways to establish whether a student actually understands what they have submitted. But that approach is difficult to implement when hundreds of students are enrolled in a single course.
Prof. Ami Moyal, chair of the Planning and Budgeting Committee of Israel's Council for Higher Education (CHE), is one of the pioneers of artificial intelligence in Israel. He sees AI as an opportunity rather than a threat.
One of his central initiatives since taking office about a year ago has been securing large dedicated government budgets to give researchers access to supercomputers capable of running AI systems and analyzing large amounts of data. He has also established several committees within the CHE tasked with preparing academia for artificial intelligence.
"AI won't replace employees with a strong AI orientation. It will replace employees who don't have that orientation," Moyal says.
He argues that universities therefore need to change the profile of their graduates so that they leave with not only knowledge but also skills such as critical and creative thinking, teamwork and the ability to present results, all of which are essential for working effectively with AI.
2 View gallery
Prof. Ami Moyal (left), Prof. Manuel Trajtenberg and Dadi Perlmutter. Universities will need to update their curricula every year in line with technological developments.
(Photos: Alex Kolomoisky, Orel Cohen, Gil Nechushtan)
"The way to train them may not necessarily be through high-enrollment introductory courses, but rather training throughout the regular courses across the whole degree," says Moyal. "Lecturers will need to focus on discussions instead of lectures, initiate exercises, guide, mentor, and be genuinely good teachers."
The CHE currently invests NIS 100 million in training academic staff, while the institutions themselves invest another NIS 100 million. Moyal is considering increasing the budget. The problem is that, from the perspective of academic staff, devoting more attention to teaching can come at the expense of research.
Prof. Manuel Trajtenberg, former chair of the Planning and Budgeting Committee (Vatat) and former chair of the Institute for National Security Studies, notes that attendance at traditional lectures had already been declining before the arrival of generative AI.
"The big change was during Covid," he says. "I expect students will come to class to hold discussions, because there's no substitute for that. But that needs to be done in small groups."
The difficulty is that a large portion of university education still takes place in large introductory courses.
"AI is good at telling you what you want to know, but it doesn't help you internalize it, and that's a big difference," Trajtenberg says. "It doesn't help you be critical and produce insights. That requires interaction, which is why discussion groups are needed."
Until now, universities have relied heavily on written papers and exercises to train and assess students.
"One question is whether you can still ask students for papers, especially in advanced degrees, when the first thing they do is turn to AI," says Trajtenberg. "True, there are applications that check whether something was written by AI. But it's an arms race between AI and those applications, and no one will win it."
That leaves universities with greater reliance on oral exams, conversations and presentations. But, again, that approach does not work easily in mass higher education.
"It's an unsolved problem," Trajtenberg says. "In addition, if the goal of training is to impart skills, you need to know how to assess mastery of them, and that's much more complex than knowledge-based exams."
"Universities understand that teaching needs to change," says Dadi Perlmutter, chairman of the Technion's executive committee and chairman of the Committee for Human Capital in Hi-Tech. "The question is how you give a student skills. It's clear that won't happen through a frontal lecture with the lecturer standing at the board and students taking notes. Lecturers will have to change. It's a new world, and the first step is convincing everyone that we need to go through this process."
The transformation is already affecting what students choose to study.
Over two years, the number of students beginning computer science studies fell 15%, from 7,277 at the start of the 2022/23 academic year to 6,219 at the start of 2024/25. The decline has been attributed to the junior crisis in software development and fears that artificial intelligence will make many workers in the field redundant.
At the same time, a significant portion of the students who moved away from computer science appear to have shifted into related fields considered more promising in high-tech, including data science, data engineering and artificial intelligence. The number of students beginning these fields more than doubled over the same period, from about 500 to about 1,042.
Over the past year, the CHE approved nine institutions to open degree programs in artificial intelligence, potentially increasing the number of students entering the field. But Moyal believes it is too early to write off computer science.
In his view, the field also needs to update its graduate profile. Software developers will increasingly need to manage teams of AI agents capable of writing software. In other words, the importance of writing code itself may decline, while defining tasks, directing AI systems and monitoring their results become more important.
Perlmutter adds that "the academic degree is still relevant and important. Most of the hi-tech industry in Israel is still interested in degrees. Basic knowledge also remains important, because you wouldn't take someone who knows how to use Claude and let them develop a missile. In computer science there was a decline, but in engineering studies the departments are bursting at the seams, and the army is looking for physicists."
In his view, institutions will need to update their curricula every year in response to technological developments. Degrees will also need to include far more applied projects and practical, hands-on experience for students in industry.
The changes are not limited to teaching and assessment. AI is also altering how academic research is conducted.
Trajtenberg sees AI as an exceptionally powerful tool for researchers.
"It's like a research assistant who works 24/7 and doesn't complain," he says. "It's amazing how much it helps. I just did something in half an hour that I would once have had to ask a research assistant to do, and it would have taken them half a day."
But he does not believe this makes researchers redundant.
"That said, artificial intelligence won't make the researcher redundant. Even when research is done with the help of AI, the researcher writes the prompt and presents the problem and the directions for a solution. But it's worth remembering that when it comes to AI, we haven't seen anything yet."
