MIT says AI can do most undergrad work, and the school is changing how it tests students.
The report from MIT's Ad Hoc Committee on AI Use in Teaching, Learning and Research Training lands with a simple claim that stops a lot of people mid-scroll. "Artificial intelligence can now handle most undergraduate assignments." That line is not about cheating. It is about the fact that essays, math problems, proofs, and coding tasks can be done at a sensible level by generative AI.
The temperature online splits fast. One side is angry and tired. They see years of late nights and problem sets turned into something a machine can draft while you grab coffee. The other side is calm and practical. They say this was coming, and the only honest move is to change how learning is proved. A smaller group goes quiet. They are the ones who already use AI to study, and they know the ground has moved under them.
MIT's answer is not to ban AI. It is to redesign the game. "The findings call for MIT to review most of its classes." The committee wants instructors to rethink learning goals and to use assessments that are harder to fake with a chatbot. Think oral exams, semester portfolios, project presentations, and in-person conversations about work done outside class.
This is not a small tweak. "Most classes taught at MIT will require careful review and, in many cases, substantial adaptation to survive an encounter with AI without serious disruption," the report says. The school also wants every course to state clear AI rules. Some assignments may allow AI. Some may require it. Some may forbid it.
The human part of this is where the report gets blunt. It notes more isolation on campus and weaker lines between students and instructors. Office hours, study groups, and undergraduate research feel different when a model can draft a first pass of almost anything. The fix, in the report's view, is to lean into people and place. "Add more physical spaces for students and faculty to meet face-to-face." The idea is simple. If AI can do the paper, then the value shifts to the room where you explain your thinking and get pushed on it.
MIT President Sally Kornbluth frames the moment in stark terms. "This is a watershed for MIT, and for all of higher education." She also makes it clear this is not optional. "This is not an optional exercise." The school is asking for AI-aware curricula, stronger residential life, and ongoing teams to keep updating policy as the tech changes.
The internet reacts to that line in a few predictable ways. Some read "watershed" and roll their eyes, sure this will end in a dozen pilot programs and a new task force that meets once a month. Others take it as a rare honest admission from a top school that the old model is cracked. A third group treats it as a signal to double down on skills that do not compress into a prompt. Build, test, present, argue, ship.
There is a real fear under all of this. If AI can do most undergraduate assignments, what does the degree measure? The report does not claim to have the final answer. It pushes MIT to ask what role grades play and to test other ways to show mastery, including competency-based systems and percentage scales used in other countries. It also warns against AI detectors, calling them unreliable and harmful, especially for multilingual students.
The unresolved question is not whether AI will be used. It will. The question is whether schools can make the human parts of learning matter more, not less. That means more time in rooms where ideas get challenged, more work that requires judgment, and more proof that a student can think under pressure. It also means accepting that some old assignments are done as a measure of skill.
For students, the shift is personal. You can feel it in the way you study now. You can generate a draft, then spend your time on the parts that require you. The risk is that you never learn the basics because the model hides the gaps. The opportunity is that you learn faster, if the course is built to force you to show your work in ways AI cannot fake.
For faculty, the load changes. Less time grading essays that look similar. More time in conversation, in labs, in design reviews. The report calls for new support for instructors, including training, AI leads, fellows, and a pilot fund to try new approaches. That is the part that will decide if this is a real shift or a memo.
For everyone watching, the MIT report is a warning shot. Other schools will face the same pressure. Some will move fast. Some will wait and hope the problem fades. It will not. The tech will get better. The assignments that rely on text output will get weaker as proof of learning.
The path forward is not romantic. It is practical. Build more spaces for people to meet. Make every course state how AI can be used. Test students in ways that require them to think out loud and show their process. Keep the parts of education that make humans better at hard problems, and drop the parts that a model can do while you sleep.
That is the feeling after reading the report. Not panic. Not hype. A clear sense that the center of gravity has moved, and the only honest response is to move with it.
More on: Entertainment