By Layla Zulfa 
In the hallowed halls of the University of California, Berkeley, the traditional architecture of mathematics education is beginning to crack. Richard Borcherds, a distinguished professor and recipient of the Fields Medal—often described as the "Nobel Prize of Mathematics"—has made a decision that signals a seismic shift in academia: he has stopped grading his students’ homework for accuracy.
For Borcherds, the decision was not born of apathy, but of a pragmatic surrender to the inescapable reality of generative artificial intelligence. After observing a surge in students submitting assignments that were flawlessly solved by AI, he concluded that the traditional model of paying teaching assistants to grade homework for correctness had become a futile exercise.
This development arrives at a moment of existential crisis for the field. Following OpenAI’s recent, landmark announcement that AI agents had successfully tackled the Navier-Stokes existence and smoothness problem—a monumental "Millennium Prize" challenge that has eluded human mathematicians for decades—the academic community is facing a reality that once seemed confined to the realm of science fiction.
The Chronology of a Paradigm Shift
The erosion of the traditional homework model did not happen overnight. It was a slow, creeping process that accelerated alongside the capabilities of Large Language Models (LLMs).
- The Early Warning: Roughly five years ago, the mathematical community largely viewed AI as a tool for computation or data organization, rather than a threat to the core of creative mathematical thought. At that time, mathematics was considered one of the "safest" careers, protected by the need for deep, abstract logical deduction.
- The Proliferation: As GPT-4 and subsequent models became capable of writing coherent, logically sound proofs, the sanctity of take-home assignments began to dissolve. Professors across Berkeley began noticing a pattern: students who previously struggled were suddenly submitting perfect, high-level mathematical arguments.
- The Breaking Point: The catalyst for Borcherds’ public admission was the recent breakthrough regarding the Navier-Stokes equations. When AI solved a problem that has defined the careers of the world’s greatest mathematical minds, it effectively signaled that the barrier to entry for complex, original proof-based mathematics had been permanently lowered.
- The Current State: Today, Borcherds has shifted to a "completion-based" grading system. He acknowledges the imperfection of this approach, admitting it is a "not a very good system," but he argues that it is the only logical path forward in a world where students can outsource their cognitive labor to machines.
The Impending Obsolescence of Human Mathematicians
The implications of Borcherds’ perspective are chilling for those who have dedicated their lives to the discipline. He posits that if current growth trajectories in machine learning persist, the window for human intellectual superiority in mathematics may be measured in years, not decades.
"If AI continues to improve at its present rate, then, in a few years, there will be nothing human mathematicians can do that AI can’t do better," Borcherds stated during a recent interview.
This creates a paradox for educators. Borcherds, a world-class mentor, finds himself unable to provide the most fundamental piece of advice to his students: whether or not they should pursue mathematics as a career. He even confessed to a surreal moment of modern irony: he asked ChatGPT for advice on how to guide his students, to which the AI candidly replied that it, too, had no idea what the future held for the profession.
Historical Precedents and the Weight of Inertia
Borcherds draws a somber comparison between the current state of mathematics and previous industrial upheavals. The transition from hand-weaving to the automated loom and the replacement of the horse by the motor car serve as warnings for the modern academic.
However, he notes a critical distinction: "This is the first time anything like this has happened in mathematics." While previous revolutions replaced physical labor, the AI revolution is encroaching upon the highest level of abstract, human cognition.
Despite this, Borcherds sees a profound level of "inertia" within the university system. He expects that even as the foundation of the field shifts beneath their feet, academic departments will continue to function as if the status quo holds, much like how print journals continue to exist despite the digital age having rendered them largely obsolete for decades. He likens the current state of math departments to these journals: functioning on tradition long after the utility has evaporated.
The Official Response: A Fragmented Strategy
The University of California, Berkeley, is not standing idly by, though its response is marked by the complexity of institutional governance. Anne Marie Richard, the assistant vice provost for undergraduate education and chief academic technology officer, notes that the university has moved away from a "one-size-fits-all" approach to AI.
The Role of the AI Hub
The administration has launched the "AI Hub," a centralized repository for tools, training, and policy samples. The current directive from the Berkeley Division of the Academic Senate is that faculty must include a clear, written statement in their course syllabi regarding the use of AI. However, there is no centralized mandate on what those policies should be.
"Direct policies on AI use can vary dramatically with course level and focus area," Richard explained. This decentralized approach allows for pedagogical flexibility but exacerbates the confusion for students who face vastly different rules from one classroom to the next.
The Failure of Detection
A significant hurdle for the university has been the search for a reliable policing mechanism. Berkeley has experimented with various AI-detection software, such as Turnitin and Rumi, but the conclusion has been unanimous: these tools are insufficient. The university has determined that they are not accurate enough to be standardized, leaving faculty to rely on their own intuition or, like Borcherds, to move away from assessment models that rely on high-stakes, unsupervised work.
Implications for the Future of Academia
The ramifications of this shift extend far beyond UC Berkeley. If a Fields Medalist—a pinnacle of human achievement in the field—cannot see a path forward for traditional assessment, the implications for the broader education sector are profound.
1. The Death of the "Homework" Era
The traditional model of education, which relies on homework as a primary assessment tool to ensure students have practiced and mastered the material, is likely nearing the end of its life. If the assignment is to solve a problem, and the student can procure the answer in seconds, the pedagogical value of that assignment has shifted from "learning to solve" to "learning to prompt."
2. A Shift toward In-Person, Oral Assessment
To combat the loss of academic integrity, universities may be forced to revert to archaic or highly resource-intensive forms of testing, such as oral exams or in-class, handwritten assessments. These methods, while historically sound, are difficult to scale in a modern university system with thousands of students.
3. The Re-evaluation of Mathematical Value
If AI can produce proofs that are superior to those of humans, the value of the human mathematician may shift from "problem solver" to "problem proposer." The mathematician of the future may not be the one who writes the code or the proof, but the one who determines which problems are worth solving and how to ethically implement the solutions.
4. Psychological Toll on the Academic Class
Perhaps the most overlooked consequence is the psychological impact on the faculty. The sense of "inertia" that Borcherds describes is not merely a bureaucratic issue; it is an existential one. Professors who have spent decades mastering a field are now grappling with the realization that their expertise is no longer the bottleneck of scientific progress.
Conclusion: A Crossroads for Humanity
The situation at Berkeley serves as a microcosm for the global academic community. As AI continues to bridge the gap between simple task automation and high-level creative synthesis, the university—a place long considered the final bastion of human intellectual pursuit—is being forced to redefine its very purpose.
Richard Borcherds’ quiet revolution—the decision to stop grading—is not just an administrative adjustment; it is a profound admission that the world has fundamentally changed. As the Navier-Stokes breakthrough proves, the machines are no longer just assisting; they are leading. The question for the next decade will not be how we prevent AI from doing the work, but rather, what work is left for the humans once the machines have mastered the art of the proof.
Sagoh 
Pevita Pearce 
Ammar Sabilarrohman 