AI was great until it was not

TEXT START: When I arrived at Oxford, I experienced what many would call a cultural shock.

The Dissection

The essay is a firsthand autopsy of intellectual community after AI enters an already damaged academic system. It documents the replacement of reciprocal learning with answer retrieval, mentorship with chatbot mediation, and scientific debate with automated validation and synthetic verbosity.

Its strongest observation is that AI did not create the loneliness, status games, weak pedagogy, or performative publishing. Those conditions already existed. AI industrializes them. It gives isolated students an endlessly available tutor, gives insecure researchers a compliant brainstorming partner, and gives competitive academics a machine for producing plausible-looking output at scale. The result is not collaboration upgraded. It is collaboration hollowed out and wrapped in efficiency rhetoric.

The article also reveals the first layer of productive-participation collapse inside elite knowledge work. The author can still learn from AI, but the institution increasingly asks a more dangerous question: if a machine can perform much of the visible work, why maintain the human apprenticeship pipeline?

The Core Fallacy

The central error is attributing the damage primarily to human misuse: “It’s not AI’s fault, it’s ours.” Under the Discontinuity Thesis, that distinction is strategically irrelevant. Competitive systems reward the actor who uses the cheaper, faster cognitive substitute, regardless of whether the substitution degrades education, mentorship, or truth-seeking.

The essay treats humane collaboration as if it could remain the governing model once AI makes solitary production cheaper and more scalable. It cannot. If one researcher with AI can generate, test, summarize, code, and publish more output than a team of human apprentices, institutions face relentless pressure to reduce the team. The disappearance of meaningful collaboration is not a cultural accident. It is the social consequence of cognitive labor becoming reproducible infrastructure.

The author also conflates intrinsic scientific value with economic protection. Understanding, mentorship, and truth may remain valuable to humans. That does not preserve the jobs, funding, status hierarchy, or number of students required to produce them. The market does not defend beautiful intellectual processes merely because they are beautiful.

Hidden Assumptions

• That academic institutions can voluntarily reserve meaningful work for humans while competitors automate it.
• That human error is uniquely disqualifying, while AI error can be controlled without changing the production model.
• That faster publication is merely a corruption of science rather than a predictable output of competitive automation.
• That PhD students will remain necessary because they provide intellectual development, even when institutions primarily purchase research output.
• That the social function of collaboration can survive after its productive function has been automated.
• That better professors and kinder peers could restore the old environment without confronting the underlying incentive structure.
• That “truth-seeking” is powerful enough to defeat citation races, labor substitution, funding pressure, and institutional cost minimization.

These assumptions preserve the moral language of the old order after its material foundation has begun to fracture.

Social Function

Primary classification: partial truth and transition management, with elements of ideological anesthetic.

The essay is not empty copium. It accurately identifies a real loss: AI can make intellectual work more efficient while making intellectual life less dialogical, less accountable, and less human. It also correctly sees that AI amplifies preexisting academic pathology.

But its final framing softens the mechanism by relocating responsibility to “how humans utilize” technology. That turns a structural transition into a character problem. The competitive system does not need malicious users. It only needs users willing to produce more cheaply and quickly than those who refuse. The essay mourns the corpse while still asking whether better manners could revive it.

The Verdict

This is a perceptive grief document written from inside the first breach in elite cognitive labor. It correctly diagnoses the erosion of mentorship, peer learning, and scientific companionship, but misidentifies the cause as optional misuse rather than compulsory competitive substitution.

The chatbot is not destroying science because it validates people. It is exposing that much of the institution already valued throughput, prestige, and plausible output over understanding. AI removes the remaining disguise.

Under the DT framework, the author’s fear is not that AI will make science less fun. The terminal threat is that science will no longer need a mass population of human apprentices at all. A small Sovereign class will control the models, compute, data, and research direction; a shrinking Servitor class will maintain the physical and institutional infrastructure; everyone else will be economically ornamental unless transfers or niche usefulness preserve them.

The essay sees the cultural death clearly. It does not yet accept the mechanical death underneath it.