What is Human good for?

After a rather heated conversation with my friend regarding AI and education, I started thinking more deeply about what humans are good for. Knowing that kids in middle school and high school have essentially stopped “learning” since they can bypass all the assignments and exams with AI, it inevitably gets us scared. Yet bypassing such assignments and assessments seems to be the steepest gradient descent of human behavior. We might have to let go of our pride and admit that we are lazy and aren’t willing to put in the hard work.

But then, why is learning hard in the first place? Why is it hard to do hardcore studying, when we are trying to learn real analysis, PDEs, signal processing — all the algebra that would drive us quite crazy? I guess it is natural at this point in human history to say that a college student would have to study hard for such subjects. But if we extend our timeline, even just a hundred years ago, we weren’t that good at reading as a society. The percent of people that were literate in 1900 is 21%. Most people don’t know how to read even today. The majority of humans actually do not know how to do algebra. They do not understand the basics of physics. Yet we live our lives. Funny enough, from a model standpoint, there is also this phenomenon, Moravec’s paradox, where it is the opposite of what we think: what is hard to learn for humans is easy to learn for models and robots.

I know if you are reading this in 2026, you will not like this. Maybe humans were not designed to read, process language, and do math. It is hard to find the exact time when humans started using written language, but it is not that long — well, not long enough to make a dramatic impact on our evolutionary history. And if the AI revolution is truly as revolutionary as some industry leaders claim it to be, it might just be the case that the industrial revolution taught humans to read and do algebra, and the AI revolution makes humans forget them.

Then what are humans good for? Or what is our current carbon-based being good for, compared to silicon-based counterparts? From my perspective, we have an incredible sensory system that is interconnected, which generates emotions and intuition. We are also infinitely self-perpetual and self-driven. And most curiously, our biological system itself is entropy-negative. To better appreciate these qualities, let me give some examples. As a trail runner, I enjoy running downhill on a rocky road, though I am far off from being the best runner. My ability to coordinate my legs and body and balance is orders of magnitude better than current robots, both in terms of energy efficiency and adaptability. We are also superb at giving ourselves purpose and being nimble with purpose. As much as there is a prevalence of mental disorder nowadays, we still live with purpose, and we can literally make up a purpose, and as we talk to ourselves about this purpose, we start believing in it — like life has a meaning at all. I genuinely do not know how to write a reward function that robust.

Maybe, just maybe, from a longer time horizon, we have overlooked the ability to read and do logic deduction in the current century. And maybe, just maybe, we are not that different from animals.