AI in higher ed: just like the shock of the Internet and the web?

AI is impacting higher education. What historical frames of reference are useful for imagining the contours of the consequences? This blog post builds on the accumulation of my postings in the last 4 years on the meanings and consequences of gen AI, right from a note on the significance of ChatGPT when it was released in 2022. I am interested in a broad question — namely, the impact and consequences of AI on higher ed. This blog post is a short side quest, where I first establish whether AI can be compared to previous tech shocks that impacted higher ed, and how so. Because if AI is “just like MOOCs: vast expectations but limited impact”, then we have learned something interesting: AI is not much to bother about. But that’s not the conclusion I’ll reach ;-) Before AI, 3 notable tech shocks to higher education Since 2000, higher ed has had its fair share of shocks that called for urgent adaptation: The Internet and the web The web brought digital content production to independent individuals — everyone got a voice and access to a potential worldwide audience. With YouTube, everyone could record a lecture and broadcast it. Learning habits diversified with shorter attention spans, screen-first and mobile-first habits. One question at the time was: would these changes sideline higher education organizations, since knowledge had become available for free, easily and from anywhere? Big data and data science Big data and data science opened up the prospect that personalized learning could replace standardized teaching at the group level. Through the analysis of vast quantities of personal data, individual learning paths could be identified. Algorithmically, or through machine learning, the adequate pedagogical resources could be identified to fit each student’s needs, progress rate and aspirations. MOOCs MOOCs promised to leverage the first two items (the web and big data) to transform education. The web offers online teaching to a limitless audience, while big data and data science allow learning paths to be individualized, keeping individuals and a tailored pedagogical experience at the center despite massification. A few companies tried to flesh out the promise: Coursera, edX, Udemy, Udacity, Khan Academy, … The question at the time was: would universities be able to survive if online equivalents existed, available 24/7 and delivering certificates from Ivy League institutions at a fraction of the price and cost of a brick-and-mortar university? Consequences of these technological shocks: much ado about nothing I am deliberately painting it in broad strokes: The Internet and the web left the core of the pedagogical experience surprisingly untouched When compared to the direct, destructive effects the web continues to have on cultural organizations (the news media, the movie industry, bookstores, …), we can be surprised that higher education remains quite stable and unscathed by the web and the digital economy. The core experience in higher education remains offline: courses taught by a professor in a classroom to a group of students. Said differently: schools and universities have been transformed by web-era technologies (email, learning management systems, video conferencing, online recruitment systems, etc.), but these have transformed operations more than the basic classroom format. Even touch-enabled smartboards (when present in a classroom) are used with moderation, in my experience. In sum, the web has been “absorbed” as yet another topic to be dissected in the classroom — rather than transforming the classroom. Big data and data science have led to the creation of specialized courses and programs From around 2015 onward, most schools started developing programs offering a crossover between [name a traditional domain] x [big data / data science / analytics], just as they had introduced classes in [digital] x [name a traditional domain] a few years before. This is a consequential change for sure; however, it did not modify the core missions or functions of higher education organizations. This is quite “normal” and underwhelming compared to expectations about the transformative potential of big data: the promise was that it would make schools capable of designing individual learning paths thanks to data analytics on student data. This promise has not been delivered at the scale once imagined. Learning analytics and adaptive tools exist, but they have not replaced the basic group-based structure of higher education. MOOC platforms still exist, but schools and universities are fine Coursera, Udemy, Udacity, edX and Khan Academy are still around, after many difficulties and restructurings. They did not displace higher education organizations by any means, but instead address new, different, or overlapping but not fully coextensive segments of students. Distance learning is a central feature of MOOCs and is indeed important, even vital, to higher ed, but that was revealed by the COVID pandemic (2020-2022) more than by MOOCs. Traditional higher education organizations adapted to the pandemic swiftly by accelerating their investments in digital services (with Zoom having had its IPO in 2019 — lucky timing). I experienced the shock of the emergence of the web in higher ed as a student, and I was an active participant in the two other shocks as a professor and program manager at emlyon business school from 2014 onward, initially under the leadership of Bernard Belletante. Belletante is a visionary who anticipated these shocks and made sure they were translated into the programs and support services of the school. The launch of new programs in data science, the creation of a Makers Lab, a new LMS, the de-siloing of academic departments, the recruitment of professors with new profiles, the adaptation of classrooms for hybrid learning, etc.: the school served students well by making all these changes in advance rather than in reaction to the shocks. Among these three shocks, the development of the Internet and the web remains the one I would personally be most tempted to compare with AI, given the transformative effect the Internet and the web had on society in general. AI is set to have an impact at least as big. And so, given the relatively weak impact the web has had on the core experience of learning and teaching (as discussed just above), a comparison between AI and the web could be illuminating. Maybe AI will turn out to be as important as the web at the societal level, but with similarly “weaker than trumpeted” transformative effects on higher education itself? Focus on the web as a historical frame of reference: is AI the same kind of shock to higher education? Let’s state what I consider to be indisputable facts. The impact of AI is: profound: On a growing number of benchmarked cognitive tasks, frontier AI systems now meet or exceed human baselines. broad: usage is widespread at school, in the workplace, in our personal lives, in governments and administrations, in arts and culture, in science, medicine and technology, in the conduct of war, etc. systemic: AI is creating or intensifying environmental, geopolitical and societal imbalances. rapid: it is July 2026 and ChatGPT was released in November 2022. So it all happened in less than four years, and the rate of development is accelerating, not slowing down. Let’s compare AI and the web: how do they compare in terms of the four dimensions laid out above? Profound shock: web and AI alike? The Internet and the web were definitely “profound” in nature, because they opened a space where new content and experiences could be created with low barriers in terms of cost, distance or authority. This created an explosion of digital services and transformed the offline world. But AI is a layer deeper. It does not merely expand the space where humans can express and develop their creativity, as the web did: it expands creativity itself. Let’s develop this idea by coming back to higher education: the web has offered students access to new kinds of resources for learning, which made it easier for them to develop the corresponding skill. AI goes much deeper: AI can readily generate in a few minutes what a skilled student would have created in a few hours or days. This is not about offering more space for expression, or connecting spaces. It is about fundamentally changing the meaning of “expressing oneself”. One can then legitimately wonder: what is the value, for a student, of learning and acquiring the skill? This question was not opened up in such a radical way by the emergence of the web. Broad shock: web and AI alike? The web can be said to be a very “broad” technology: it touches everything, in particular since we pack so much of our personal and professional lives into the smartphones we carry all day long. But there again, AI is broad at a more fundamental level. LLMs have the capacity to emulate everything we ask them to be. Just as there is a conceptual gap between digital devices and analog mechanisms, LLMs introduce a new gap that sets them apart from “traditional” digital devices. LLMs are still software, of course. But the service they provide is not akin to software that simply follows a predefined series of instructions (the argument is developed there). Instead, they can be used for any purpose we set for them at the point of use: explaining a concept from any domain to undergrad students or to PhDs, for instance; helping a professor create the content of a class; or helping program officers review entire curricula. AI is a “broader” shock than the web in this sense: the web had a broad impact, but AI is broader because it is a kind of nearly omnipotent thinking device. Systemic shock: web and AI alike? The development of the Internet and the web at scale has had profound environmental, geopolitical and societal consequences (e.g., a, b, c). Yet it seems that AI has even more consequential effects. Developing and running AI models necessitates data centers that consume large amounts of electricity, critical minerals and water. The planned data-center capacity to be created in the next decade is widely discussed as putting other uses at risk. AI models could also create unemployment among white-collar workers, and also among blue-collar workers when these AI models are used to augment robots with smart behavior, performing better and more cheaply than human operators would (taxi drivers, factory workers, …). The Internet and the web do not run in a literal cloud, and access to the Internet and the web is certainly not free. The Internet rests on a physical infrastructure of data centers, cables, satellites, internet service providers, local and global authorities… and it all continues to require massive investments and maintenance costs by schools in what was called “digital transformation”, a transformation that never seems to come to an end. Here again, AI seems to follow the same logic, but at a larger scale. To repay their investments in model training and infrastructure building, OpenAI and a few other major players in gen AI have started offering subscription services that range from free plans to a few hundred dollars per month, per individual. With the spending rate of these big players accelerating, and with the multiplication of sophisticated AI services, we can expect AI to open a new category of significant spending for schools. NB: true, open-weight models can be acquired for free and run locally, but they tend to be associated with their own specific and significant costs, notably the cost of ownership of the IT infrastructure required to run these models and the human resources in charge of installation, maintenance, security, user onboarding, etc. While the Internet remains relatively open across most borders (with major caveats: see China, Russia, etc. [Martel, 2018]), AI services are created and distributed from two main regions: the United States and China. The release and rapid shutdown of Fable 5 by Anthropic in June 2026, following a US government directive that required Anthropic to restrict access for non-US citizens, makes the dynamics of access to frontier AI services feel quite different from the early open expansion of the web. Europe has credible AI actors, notably Mistral AI in language models and Black Forest Labs in visual intelligence. But the frontier AI ecosystem remains much more concentrated in the United States and China than the open web ever was. This creates a dependency risk for European schools and universities. AI models also increase the potential for harm in cyberattacks or in the design of bioweapons (Suleyman, 2024). Finally, AI is considered by some to pose an existential risk to humanity itself, in plausible scenarios of loss of control, misuse, or misaligned objectives (Hinton, 2025, AI Statement). The web, while not a calm and quiet innovation, certainly never amounted to this level of systemic risk. Rapid shock: web and AI alike? The adoption of web technologies was relatively “rapid” on a historical timescale, in the sense that it took about two to three decades for web usage to become widespread after the emergence of the first web browser. By contrast, according to Sensor Tower data reported by Reuters, the ChatGPT app reached 1 billion monthly active users globally roughly three years after ChatGPT’s November 2022 release. Conclusion In higher education, there is definitely some truth to the Gartner hype cycle: when tech shocks occur, a peak of inflated expectations forms quickly; then the dust settles, and a plateau of productivity is reached, where universities and schools absorb the innovation while their centuries-old defining features remain completely intact: source: The Gartner Hype Cycle Maybe AI will be of the same sort. That would mean that in 10 years’ time, following the playbook of what happened with previous tech shocks in higher ed, we could expect: “Experts in AI visual creation” in schools of arts, “AI for finance” programs in business schools, AI-enabled scheduling systems used routinely by planning teams, AI avatars doing the job of admissions counselors, … and a human professor still teaching human students, exactly as has happened for centuries. But we established above that AI cannot be compared to previous tech shocks. It is more profound, broad and rapid, and it has vastly more systemic effects. For this reason, and that is where I want to stop: we can’t safely assume that AI will be as “innocuous” as the web and other tech shocks that have impacted higher ed in recent decades. From there, we can embark on the study of the contours of this impact. This is for the next post! About Me I’m an academic and independent web app developer. I created nocode functions, a point-and-click tool for exploring texts and networks. Try it out and let me know what you think. I’d love your feedback! Email: analysis@exploreyourdata.com Bluesky: @seinecle Blog: Read more articles on app development and data exploration.