1. In-class exams

The quizzes and exams may be in-class. If one heavily uses AI do homework and try to “learn” by only reading the outputs, they will certainly suffer in exams.

Math is not a spectator sport. It is impossible to learn math well only by reading. One must get their hands dirty and struggle through deep thinking, in order to truly understand the concepts, and hence excel in exams. (This is not an opinion, but a general phenomenon: Every mathematician agrees on this.)

It is like going to the gym: If you are not suffering, you are not doing it right: The suffering is the whole point. For math, the suffering is the learning process. If everything feels so easy, probably you are not learning it right.

So, if you care about GPA, then I think the above should be a pretty sufficient reason to not over-rely on AI when doing homework.

But you may ask: If AI can already do all undergraduate math (which is quite true), what the hell is the point of in-class exam? Why do I need to know all these math by heart? After all, in real life, no company will ban their employee from using AI, if AI can increase productivity.

Good question. This brings us to:

2. Job prospects

Everyone wants intelligent coworkers, who acquire clear and deep understanding of the tasks. No one likes a coworker that only have shallow understanding of everything, and is glued to an LLM all day. There are two reasons behind:

2.1. Practical

If candidate A is better than everyone else in doing a certain task, then very likely candidate A will also be better at using an AI to do the task, just as experienced coders tend to be better at vibe-coding than inexperienced coders.

Learning how to use AI is easy, once you know what you are doing.

But it is not just about productivity.

2.2. Human nature

People just like talking to smart people. Even if an LLM can answer all my questions, I would definitely still prefer chatting with other human: It’s just way more fun!

This is built into our psychology: There exists natural affinity (sometimes repulsion) between human beings, which do not exist between a human and a robot.

Now, imagine yourself entering the job market. Through in-person interviews, it is extremely easy for employers to distinguish an intelligent and thoughtful candidate who set for themself high standards, from one who is hollow inside and survived 4 years of college via abusing ChatGPT. Trust me - it is very easy to tell the difference. This is also why oral exams (though time-consuming) are widely considered much better than written exams in judging students’ ability.

We are not just tools: Employers do not only care about what we can do, but also who we are.

3. Healthy cognitive diets

(I first heard this analogy with healthy diet and life style from Terence Tao.)

As you probably have heard: Universities are not merely training centers for industry jobs: They also, among other things, foster our intellectual growth. From this point of view, what are the effects of over-reliance on AI?

Returning to our math classes: Beyond securing good grades and good careers, what are the benefits of devoting serious mental efforts to struggle through homework?

Let’s consider an analogous question: What are the benefits of going to gyms?

3.1. Gyms

100 years ago, people did not go to gyms. That’s because they did not need to. Back then most jobs are hard labor anyway: There’s no need for extra workout. But then came the industrial revolution. From scarcity in labor, we have abundance in labor, by machines. With the rise of white collar jobs, more and more people just sat in office all day, and became obese. Gradually, people realized the importance of healthy life style. Going to gym then became trendy.

The point is: With new technology comes new responsibility, in maintaining our own well-being.

3.2. Homework

With the internet and AI, we are entering the age of abundance of information and intelligence. How would that impact our brain?

Well, what have Youtube, Instagram, and Tiktok done to our brains? Everybody knows: They feed on our impulse to fulfill momentary visual pleasure, and have thereby greatly decreased our attention span. Nowadays reading novels, or even watching movies, are considered demanding!

With new technology comes new responsibility. I argue that, the intentional act of doing mental hard work, e.g. homework, is an ideal way to defend and train ourselves.

It is very tempting to use AI to solve all homework problems. In short term, this may appears to be an economical way to earn good grades. But in doing so, we are forfeiting the opportunity to develop careful and logical reasoning skills, the patience and persistence for solving difficult problems (not just in math), and even the self-discipline required to succeed, in life.

Homework should be viewed as a vital part of a healthy cognitive diet: It prepares us for other (more meaningful) tasks in the future.

One might naively reassure themself, “I will just use ChatGPT for homework. And I promise, later when it comes to things that ‘truly matters’, I will put in real effort.”

Regrettably, our brains do not function like that. We are all shaped by our habits. To be lazy for 4 years and then miraculously become self-disciplined overnight is simply psychologically (or even neurologically) unfeasible.

4. Knowledge as a path to happiness

To me, this final reason is the most important one.

The great mathematician Dennis Sullivan once commented: “What’s the secret to happiness in life? Two words: Be interested.”

In other words, stay curious, like a child. And I believe, learning is the a very effective way to achieve that.

But this deserves another long discussion: See a previous post of mine.