TL;DR
- Skill-oriented, credentialing-as-a-service higher-ed is a product of the industrial era’s need to supply and certify workers for increasingly specialized jobs.
- AI undermines higher-ed’s business model: it can perform many of the skills universities teach, and it erodes degrees’ power to certify who’s actually capable. Universities aren’t positioned to adapt quickly.
- This should be viewed as an opportunity: with AI handling skill acquisition and certification, higher-ed can return to what it was arguably always supposed to be — cultivating intrinsic curiosity and human connection.
In the movie A Beautiful Mind, John Nash (played by Russell Crowe) would mutter the words “eager young minds” as he walks into classrooms. Despite the portraying of Nash’s resentment for teaching, this scene from the movie shaped my early imagination of higher education – an “Academy” where intensely curious, intrinsically motivated students gather to learn from the best minds of their fields, and engage in the creation and dissemination of knowledge for the benefit of humanity.
The longer I work in the U.S.1 higher-ed institutions (and business schools in particular), the more I realized how rosey and idealistic the above image had been. A U.S. higher-ed institution is, at the end of the day, a business that offers skill credentialing as a service. Absent a more accurate signal, University degrees and credentials become proxies for individual capabilities, which various markets (and the society in general) rely on for resource allocation (jobs, status, and “opportunities” in general). This characterization, while may be unpopular, is far from novel or controversial (see The Abundent University by Michael D. Smith for much more eloquent argumentations).
Now, AI has fundamentally challenged this business model. When most of the concrete skills we teach can be done by AI (not perfectly but comfortably above the proficiency of an average student), the credentialing power of Universities is significantly diluted. This is because a 4.0 GPA can no longer reliably distinguish between a capable student vs. someone with a $20/month AI subscription. Meanwhile, just like large corporations troubled by bureaucracy and resistance to change, it is unrealistic to expect Universities to adapt fast enough. Taken together, it is my personal belief that AI will disrupt today’s higher-ed system, and there is very little I or anyone else can do about it. This doesn’t necessarily mean that Universities would just cease to exist, but they would take on a very different form in the post-AI era.2
The Industrialization of Education and the AI Disruption
Despite their prevalence today, the skill-oriented, credentialing-as-a-service education programs are mostly an invention after the industrial revolutions. As jobs became increasingly specialized “bundles of tasks” (e.g., Acemoglu & Autor 2011), higher-ed evolved into both a supplier of individuals who have the skills to perform those tasks and a certifier of their capabilities. This pattern is especially pronounced in professional graduate programs. For example, large industrial organizations need managers, so business schools bundle management-related courseworks into MBA programs; complex data analyses need data scientists, so a mixture of statistics, computer science, and management gives birth to various business analytics programs. In both cases, universities then provide skill certification through degrees, rankings, and transcripts.
However, AI has now fundamentally disrupted this value proposition. Frontier AI models today can ace college-level exams in STEM, humanities, law, medicine, and social science (e.g., MMLU-Pro benchmark). Naturally, students are turning to them for help when the traditional classrooms fail to deliver personalized education and support (although the impact of AI adoption on learning outcome is not clear cut). In other words, AI directly challenges the two functions that university education is supposed to fulfill: transfer of knowledge/skills and certifying who are better at those.
In the face of such a disruptive force, how have universities reacted? Well, not much. Expecting higher-ed institutions to rapidly adapt is like asking 1975’s Kodak to “just make more digital cameras”. Progress in AI is measured in months if not weeks, to the point where even the best find it hard to keep up. In contrast, in many higher-ed institutions, assembling committees to “discuss about AI” would already take that much time. As such, it is impossible for the current higher-ed institutions to “just adapt to AI”.
The Purpose of College Education
Now that we know we will get disrupted, we can go back to first principles and ask the following question: what’s the point of college education? Renowned UChicago sociologist Andrew Abbott drew the following conclusion in his 2002 Aims of Education Address:
In summary, from a practical point of view there is no evidence that undertaking the particular intellectual exercises we set for you here at college has any exclusive connection with your worldly success or your cognitive development. Nor is there really an effective theoretical argument for aims of education going forward into the future. The reality is that education is a present quality of the self, a way of being in the moment. And that quality is its own aim, because it expands our present experience and hence is worthwhile in itself.
His concluding remark summarizes this position even more succinctly – “There are no aims of education. The aim is education. If—and only if—you seek it … education will find you.” I find this to be breathtakingly illuminating. The purpose of education is education itself, and one needs intrinsic motivation (rather than extrinsic “utility”) to truly pursue it.
I argue that the AI disruption, viewed in this light, looks less like a challenge than an invaluable opportunity, for higher-ed to finally return to what it should be – inspiring people to discover their intrinsic passion and setting them on the paths to pursue it – from the utilitarian distration of mechanistic skill transfer and certification in the past several decades. Let (human) educators inspire, motivate, debate / share perspectives with students, and let students use AI to acquire concrete knowledge and skills.3
To clarify, general-purpose chatbot AIs are not good educators out of the box, but they are very good at learning from the best educators, e.g., Richard Feynman, Steven Strogatz, Grant Sanderson (aka 3b1b), Gilbert Strang, and many others. Distilling the best educators can enable the provision of high-quality education at scale. Therefore, effective scaling of best educators is a worthy technical problem to solve (see Eureka Labs).
There and Back Again
In the age of AI, major in being human. When AI can do most of the things we teach in universities, that means universities need to change, from institutions that offer degrees for fees to places where people learn to be with themselves and with each other. To me, this is a beautiful return of the value of education, and I embrace it with all my passion.
Footnotes
The “U.S.” qualifier is here simply because I have not worked in a University outside of the U.S. for an extended period of time.↩︎
All opinions in this essay are my own, and do not reflect the positions of my colleagues, the Carlson School of Management, or the University of Minnesota.↩︎
AI-facilitated training and skill certification is, in itself, an interesting research and business problem. As mentioned before, a job can be thought of as a bundle of tasks, and a degree can be thought of as a certificate that an individual can perform a bundle of tasks. With AI being able to perform more and more tasks, I expect traditional hierarchical organizations to give way to more flexible, task-based organization mechanisms. This will lead to un-bundling of jobs and the need for atomic skill training and certification with AI.↩︎