Good news everybody! Another piece of the tired 20th century may finally be on its deathbed. The university system is in terminal decline and I couldn’t be happier.
The news comes from a family member who teaches about artificial intelligence at one of the universities. In perhaps the juiciest of ironies, professors have given up trying to detect if the students are using AI to write essays and pass tests. You would think that professors of AI would be able to outrun their students, but you would be wrong. It’s over. The students have defeated the teachers.
On a certain level, given the subject matter, you could say the only test that matters when learning about AI is whether the student understands how it works and presents creative ways to deploy it. Sure, it must be frustrating for the professors to be surrounded by cheaters. But, to be fair, AI was built to improve a human’s ability to perform tasks. It is value-neutral, like a tool. Whether a tool is used for good or evil is entirely up to the user. So, pass, I guess?
If the issue of AI cheating was confined to the technical schools, that might be a sound argument. The problem is that these tools are everywhere now, in all schools and all degrees, including PhD-level courses. No teacher or professor, anywhere on the globe, has a reliable way to check if a student has used AI to pass a test or exam. If a student knows which programmes to use, their cheating is essentially undetectable.
And you might be thinking, who cares if a Gender Studies student cheats? But the scary thing is that there’s nothing stopping aspiring engineers and medical students from using the AI tools to cheat either. If you can’t see the existential crisis here, I really hope you don’t have to use a bridge or get heart surgery over the next 20 years.
Thankfully, the engineering and medical worlds have multiple, redundant layers of checks for competency. A degree isn’t a magic ticket to become a doctor. A student surgeon must go through placements, take continuing education, accrue thousands of hours of experience and their every daily action has real-world consequences in the form of blood and death. Someone who cheats on a medical exam will be found out long before they pick up a scalpel.
But if we’re being honest, the existence of AI does question the need for most doctors in the first place. After all, what is a doctor but a walking decision tree? The doctor asks the patient what’s wrong, then uses their knowledge of medicine to diagnose the illness. Yet no matter how smart the doctor may be, there is no human brain on the planet that can outperform an AI that stores the entirety of medical knowledge. I expect most human doctors, except maybe surgeons, will be among the first to disappear in the looming AI job-stealing apocalypse. Surgeons shouldn’t get too comfortable either.
So, maybe the university cheating scandals aren’t that bad after all? I mean, surely the real test going on – the meta test – is whether we can become comfortable dealing with AI for complex jobs like medical diagnoses. Because that day is coming and may already be here. If a smart machine can diagnose better than any human doctor, we should be excited about that, right? Or am I missing something?
I don’t think I am. In fact, I reckon we’re about to return to a type of guild system. Call it neo-guild. Under the guild system, prices were fixed. There was barely any such thing as a ‘free market’. The price for a carved wooden chair was the same everywhere. But there was still plenty of competition. It just happened at the level of quality. Fixed prices forced carpenters to get better at carving chairs, which led to the birth of artisanship and the production of unbelievably high-quality furniture and other items.
Something similar may be happening again due to AI. Right now, AI models are owned by private companies. But it’s not unreasonable to expect a government to eventually nationalise a mature AI database, or build its own, and deliver it as a public utility. If that happens, the price of using AI would drop to nearly $0. By plugging into the free AI, anyone could access the entirety of medical knowledge, for example, and set up their own clinic. The problem then becomes: how would a patient choose which clinic to visit? If they’re all using the same AI tool, what’s the deciding factor?
The answer would be quality. Does the doctor have a good bedside manner? Does the doctor have good intuition or judgement? Does the doctor keep up with the latest evidence? If patients retain their ability to choose their clinic, the “market” would filter out the cheaters and reward the doctors who can both deploy AI effectively and deliver the highest quality. Doctors who actually knew their stuff could then charge higher rates. This would create a positive incentive for student doctors not to cheat on their examinations and learn what it takes to become a quality doctor, since quality will be the only way to compete in a world with zero-cost AI.
Seen in that light, the AI cheating problem is simply acting as a filtering mechanism that will improve the overall outcome of medicine by forcing the unintelligent and the dishonest out of the doctor pipeline. The easy option of cheating means the real test is not whether you can pass tests, but whether you know the subject. Said differently, AI cheating tools might be giving us better doctors once the dusty old 20th century university system can rebalance their efforts to reward quality of medical service over empty degrees.
But in the coming world, what’s the point of universities? A person who cheats is only harming themselves. So why cheat at all? Did you ask this question? How about this: why were the universities so easy to fool? Answering that question reveals why universities deserve to fade away.
For a typical degree, there are no consequences for failing a test. None whatsoever. Some career paths will kick a student out if they fail too many times, things like medicine, aviation or law. However, for almost every other degree pathway, the student can fail as many times as they wish, so long as they maintain a minimum standard (very low) and are willing to pay the $4500–$6500 fee.
In other words, the incentive structure is not for students to learn anything. No university gives its graduates exit exams to measure how much they've learned since they sat their entrance exams. Rather, the incentive is for students to pass their exams using any means available, so they don’t have to pay to take the course again. In the past, that meant reading the textbooks to learn what the professor believed and repeat it back to him on the test. Now it means using AI to cheat. The university doesn’t care if students learn anything. It will take their money whether they pass or fail.
The modern university is a business that hands out defective degrees right next to legitimate degrees. And it gets away with this because society hasn’t figured out a way to objectively measure the quality of graduates. At the very minimum, universities owe students their money back, and if they don’t pay up the students should sue for breach of contract. Unfortunately, the universities saw this coming a long time ago and set up a liability shield right there in the terms and conditions on every degree: “At the conclusion of this course, students will show a proficiency in...” Who decides what “proficiency” means? Well, it certainly won’t be the student.
Under this incentive structure, you’d be stupid not to use AI to cheat. If only a “proficiency” must be shown to gain a degree, rather than a demonstrable transfer of knowledge, then the value of that degree is based on whether a graduate can trick their future employer that they are competent. Graduates who cheated quickly figure out that this is a bad strategy. You might be able to fool your engineering professor, but you won’t be able to fool that fully laden 40,000kg truck passing over your badly built bridge.
The problem is that the bridge-building company can’t simply hire engineering graduates and hope the load calculations can be done by the AI anyway. Just like in medicine, the engineering sector still requires a method to filter out who is competent and who cheated on their exams. Billion-dollar contracts and the lives of real people depend on getting those questions right. If the signals from the universities can’t be trusted, then what signals should the companies look for?
The answer is SAT and IQ tests.
The SAT is a standardised exam used to assess a student's reading, writing and maths skills, while an IQ test measures cognitive abilities such as reasoning, problem-solving, pattern recognition, memory and verbal acuity. The neat thing about these tests is that they are both reliable and don’t require a university at all. You can test children directly in high school, find the smart ones, put them on a course, mentor them to become high quality professionals and then pair them with super AI.
In the coming world, I really don’t see any room for universities and I couldn’t be happier. Don’t feel sorry for them. They did this to themselves.
