BERKELEY, Calif. — In a dramatic escalation of the ongoing friction between higher education and generative artificial intelligence, a prominent mathematics professor at the University of California, Berkeley, has taken the radical step of eliminating graded homework and online tests entirely.
Richard Borcherds, a Fields Medalist and renowned figure in the mathematical community, made the decision after concluding that the proliferation of advanced large language models (LLMs) has rendered traditional take-home coursework effectively obsolete. According to reports from the New York Post, Borcherds stated that it is no longer feasible to distinguish authentic student work from text and problem-solving generated by AI tools. Consequently, the professor has shifted exclusively to traditional, in-person examinations that strictly prohibit electronic devices, returning to a low-tech evaluation model to gauge true academic comprehension.
The move by a faculty member at one of the world’s premier public research institutions highlights a growing crisis of academic integrity. As generative AI models become faster, cheaper, and infinitely more capable of solving complex calculus and advanced mathematical proofs, universities find themselves grappling with fundamental questions regarding how learning is measured, how knowledge is tested, and whether the foundational mechanics of higher education must be entirely rebuilt.
Main Facts: The Shift Away from Digital Assessments
The core of Professor Borcherds’ decision lies in the breakdown of trust and efficacy surrounding asynchronous, unsupervised coursework. For decades, homework and online quizzes have served as the baseline scaffolding of university education—a way for students to practice concepts, make mistakes, and receive formative feedback before high-stakes midterm and final exams.
However, the rapid democratization of generative AI has inverted that pedagogical model. Tools capable of breaking down multivariable calculus, solving differential equations, and writing proof-based mathematics are now accessible to any student with a smartphone or a browser tab.
Borcherds noted the unmistakable statistical anomalies that tipped him off to the shift. In statements to the press, he pointed out that online exams and remote assignments began displaying sudden, unearned spikes in average performance.
"Online exams are now a waste of time as several students just use AI to answer the questions," Borcherds explained. "I cannot tell which students are using AI, but I notice the average score increases a lot in online exams."
Furthermore, the professor highlighted the perverse economic and ethical incentives created by grading AI-generated homework. "Grading for correctness just adds a huge incentive for them to use AI for homework instead of doing it themselves and penalizes students for honesty," he observed. "And paying to grade AI-generated answers is a waste of time and money."
By abandoning graded homework, Borcherds has streamlined his assessment strategy down to a singular, verifiable metric: the unassisted, in-person, analog test. Without laptops, tablets, or internet access, students must demonstrate their cognitive grasp of mathematics under controlled conditions.
Chronology: How AI Dismantled the Take-Home Assignment
The journey from the launch of consumer-facing generative AI to UC Berkeley’s classrooms abandoning homework has been remarkably swift, marking a breathless timeline of technological disruption in academia.
Late 2022: The Arrival of ChatGPT
In November 2022, OpenAI released ChatGPT to the public. Initially, educators viewed the tool with skepticism, noting its tendency to hallucinate facts and stumble over complex multi-step logic problems. However, within months, the underlying models underwent exponential improvements, rapidly advancing in mathematical reasoning and code execution.
2023–2024: The Era of "Cat and Mouse" Software
As AI capabilities expanded, universities initially responded by deploying digital surveillance and detection tools. Plagiarism checkers integrated AI-detection software, and online proctoring services locked down web browsers during remote exams.
However, this period was characterized by a frustrating game of cat and mouse. Students quickly learned to bypass detectors by using paraphrasing tools, running local open-source models on personal laptops, or feeding assignment prompts into multimodal models that could interpret handwritten equations. Faculty members found themselves spending countless hours investigating academic dishonesty claims, often with inconclusive results that strained relationships with students.
2025: The Tipping Point of Ubiquity
By 2025, generative AI was no longer a novel novelty; it was deeply embedded in the digital ecosystem. Integrated directly into operating systems, web browsers, and productivity suites, AI tools became frictionless. Attempting to restrict student access to AI on take-home assignments became as futile as trying to ban calculators in the late 20th century—except calculators could not generate complete, human-sounding argumentative essays or step-by-step proofs.
Late 2026: The Capitulation at UC Berkeley
By the fall 2026 semester, the friction reached a breaking point for veteran educators like Professor Borcherds. Recognizing that the administrative overhead of policing AI was yielding diminishing educational returns, and realizing that homework no longer reflected independent student capability, the decision was made to discard graded assignments entirely in favor of secure, analog testing environments.
Supporting Data: The Scale of Academic AI Adoption
The challenges faced by UC Berkeley are not isolated anomalies; they reflect a nationwide and global transformation in student habits. Educational technology surveys and academic studies paint a stark picture of how deeply artificial intelligence has penetrated the student body.
- Surging Usage Statistics: Recent surveys indicate that upwards of 80% of university students utilize generative AI tools on a regular basis for coursework, ranging from brainstorming and editing to full-scale assignment completion.
- The "Zero-Effort" Paradox: Studies tracking academic performance show a widening gap between unsupervised assessments (where students score exceptionally high via AI assistance) and supervised, in-person testing (where performance frequently dips to baseline historical averages). This disconnect suggests that high grades on homework no longer correlate with actual mastery of the subject matter.
- Resource Drain: Departments across the country report increased spending on academic integrity investigations, appeals boards, and software licenses for detection tools that independent researchers have repeatedly proven to be unreliable. Borcherds’ critique regarding the waste of time and money in grading AI outputs resonates deeply with teaching assistants and professors drowning in synthetic submissions.
Official Responses: Academia, Administration, and Tech
The reaction to Professor Borcherds’ policy shift has ignited a broader debate across the academic landscape, drawing commentary from university administrators, student bodies, and education experts.
The Faculty Perspective: Pragmatism vs. Tradition
Many STEM educators view Borcherds’ approach as a pragmatic, if severe, necessary correction. Proponents argue that mathematics, physics, and engineering education have always relied on the internalization of fundamental principles through struggle. If AI short-circuits that struggle, students graduate without the foundational competence required for advanced coursework or professional engineering.
However, other educators express concern that abandoning homework entirely strips students of low-stakes learning opportunities. Without homework feedback, students may reach high-stakes exams unprepared, leading to higher failure rates or increased anxiety.
The Student Body: A Mixed Reception
Student reactions at UC Berkeley have been polarized. Students committed to rigorous learning and academic integrity have largely expressed sympathy for professors trying to maintain standards in an unfair ecosystem. Conversely, students accustomed to leveraging digital tools for efficiency worry that high-stakes, exam-only grading regimes create an overly punitive environment where a single bad day or test-taking anxiety destroys a semester’s grade.
University Administration and Policy Makers
Administrators at UC Berkeley and peer institutions are currently scrambling to draft cohesive institutional guidelines. While some departments are doubling down on AI literacy—teaching students how to prompt, audit, and collaborate with AI responsibly—others are quietly mirroring Borcherds’ return to analog fundamentals. The lack of a unified playbook leaves individual professors to chart their own course through the crisis.
Implications: What Happens to Higher Education Now?
Professor Borcherds’ decision to scrap graded homework is more than a localized policy change; it serves as a harbinger for the future of higher education in the age of artificial intelligence. Several long-term implications are already taking shape:
1. The Renaissance of Analog Testing
We are witnessing a paradoxical reversion in educational technology. While universities spent decades digitizing lectures, moving assignments to online portals, and expanding remote learning infrastructures, the AI revolution is driving a swift retreat back to paper, blue books, and physical proctoring. Expect to see more classrooms stripped of laptops and smartphones during evaluation periods.
2. The Death of the Traditional Take-Home Essay and Problem Set
The days of evaluating students based on unsupervised written output are numbered. If a task can be completed asynchronously outside the gaze of an instructor, universities will either have to accept that AI played a role or redesign the assignment format entirely. This may lead to an increase in oral examinations, lab-based practicals, and real-time collaborative projects executed inside secure campus environments.
3. Redefining the Value Proposition of a University Degree
As AI makes information and problem-solving instantly accessible, the traditional value proposition of a university—acquiring and regurgitating factual knowledge—is fundamentally undermined. Institutions will be forced to shift their emphasis toward cultivating critical thinking, human-to-human collaboration, ethical reasoning, and the meta-skill of learning how to learn.
4. Equity and the Digital Divide
Ironically, strict reliance on in-person, high-stakes exams may exacerbate equity issues. Students dealing with learning disabilities, test anxiety, or external life pressures often relied on the flexibility of take-home assignments to demonstrate their understanding. As universities lock down testing environments, educators will need to find new ways to support diverse learning styles without compromising academic rigor.
Conclusion
Richard Borcherds’ choice to drop graded homework at UC Berkeley may have seemed like a drastic reaction to a local nuisance, but it serves as a crystal-clear reflection of the existential crossroads facing global education. As generative AI continues to evolve at breakneck speed, universities can no longer pretend that traditional grading metrics remain untainted. Whether through a return to strict analog testing, as Borcherds advocates, or through a wholesale redesign of curriculum and evaluation, academia must adapt or risk losing the integrity of its credentials entirely.
