The Clash of Wrongs

 I'd like to begin with a modified quote from Hegel:

 The real tragedies in this world often arise not from conflicts between right and wrong, but from conflicts between two wrongs.

Academia's past elitism, its cutthroat system of assigning credit, and its culture of keeping students and postdocs in poverty while forcing them into endless competition are wrong. But so are the tech bros and their accelerationist views, people don't care what happens to others, as long as they can position themselves on the side of the entrenched interests amid the potential chaos.

Part 1, I want to say something to some of the people I know working in AI

Many of you may have misunderstood the attitude of the theory (math/TCS) community. At least from what I've seen, most people in the theory community have never thought that AI producing research results was a bad thing. We simply don't think it's a good thing to keep churning out massive amounts of results without caring whether anyone can digest or understand them.

The point of mathematics is not just to arrive at conclusions. If nobody digests or understands the results, and we're simply piling up conclusions for the sake of it, then we're stripping mathematics of most of its meaning as a discipline. Unlike many engineering fields, our primary goal was never applications. Of course, if a mathematical result happens to have applications, that's great too.

But more fundamentally, we want to broaden the horizons of human thought in a philosophical sense, and keep people's curiosity about the unknown alive.

And that curiosity shouldn't stop at being intrigued by the statement of a theorem. People need to actually develop intuition and logical understanding from the process of arriving at a result. This requires time and careful reflection. Now, with machines constantly spitting out results every day and nobody having the time to digest them, it's genuinely exhausting. And frankly, AI writing is still really bad. It takes an enormous amount of time just to understand what it's saying.

Our position has never been that "academia should remain the way it was." Academia had plenty of problems long before AI came along, and now is an opportunity for change. In fact, many academics are already actively discussing new missions and directions for their fields. They're not the outdated relics or gatekeepers that AI bros imagine them to be.

But that doesn't mean OpenAI's approach of vomiting out hundreds of unpolished results has no negative consequences, or that it deserves praise. These are two separate issues and should be judged separately.

To be clear, I'm not even in academia anymore myself, but I still feel the need to speak up for it.

Personally, I genuinely don't care if AI solves a problem I'm working on. Even without AI, if someone were determined to scoop me, they could always do so. 

I also think people have a somewhat stereotyped view of the cutthroat competition in academia. At least in the theory community, most people really aren't that toxic. Sometimes I even wonder whether certain people in the AI community are projecting their own "everything is a zero-sum game" or "survival of the fittest" worldview, along with their darker assumptions about human nature, onto our community.

In reality, many people who work in theory have an almost communist spirit, contributing out of sheer passion for the subject rather than personal gain. Of course, there are some toxic individuals who care only about getting their names on as many papers as possible. But I actually think that changes may be brought to the academic credit system with AI, as the pure number of results obtained becomes increasingly worthless,  and these people may feel less rewarded with the prestige they want. Eventually, they may simply leave the community.

What genuinely worries me is the social impact of OpenAI's "brute force wins" and "money talks" style of publicity. For one thing, the careless way they dump unpolished, unreadable AI slop research results already shows that they don't care about the development of the field at all. They just want to flex their muscles.

On another note, I've always tended to side with the less powerful.

I think OpenAI's decision to focus on mathematics is, to some extent, a case of picking on an easy target. First, mathematical results are verifiable, which makes mathematics relatively easy to train AI on. It doesn't require long experimental cycles in reality such as in medicine. Moreover, the people directly affected make up only a tiny fraction of the population, and their voices aren't loud enough to draw attention to what these companies are doing wrong. At the same time, there are hardly any legal, safety, or vested-interest issues involved.

If they actually produced a result in a field that could bring about real social change and affect powerful, entrenched interest groups—say, something related to new energy technologies or new medicines—and then promoted it just as aggressively, I would have considerably more respect for them.

More importantly, what worries me even more than the chaos academia may experience during this transition is the possibility of fostering public disdain for learning, nihilism toward knowledge, hostility toward intellectuals, and eventually a complete rejection of education itself.

We learn, think, and get educated, not because we aim to become cogs in society's machine or to join the privileged class. We do so to become independent thinkers and autonomous citizens. We need to be capable of questioning AI, questioning authority, questioning the power of capital, questioning the media, and questioning governments. Mathematics is an important part of that.

Part 2: For my friends not working in AI

Will AI Actually Replace Our Jobs?

For one thing, I think those of us in STEM know very little about economics and sociology, myself included of course. I'm starting to realize that many of our worst-case fears about the future come from not understanding how certain social mechanisms work, or from not looking beyond oversimplified narratives like "AI will replace everyone's jobs." We live in an echo chamber, constantly reinforcing each other's fears.

It's not that I'm not anxious, or that I think AI won't affect jobs. In the long run, it certainly will. But I'm strongly against mindlessly spreading anxiety. Some AI bros are already fear-mongering every single day. If we keep amplifying that anxiety ourselves, I think that's honestly rather irresponsible, even inconsiderate toward others, especially the younger generation who will experience the impact most.

There are actually so many different aspects of AI's impact worth discussing, but for now I'll stick to the field I know best.

For example, I think the employment issues facing most professions are fundamentally different in nature from the issues facing academia. But I don't want to venture into areas I don't understand well enough yet. I'll come back to those once I've learned more.

Forgive me for being blunt, but I really don't think it's helpful to constantly say things like "We're doomed, everyone's going to lose their jobs" or "The party is over."

Especially when, so often, the people making these anxious statements are those who already have jobs, often senior people. Meanwhile, the fresh graduates and current students who are actually struggling to find jobs, and who may already be so much anxiety, aren't the ones I see constantly making these remarks.

Besides, looking just at the current state of mathematics and theoretical computer science (TCS), the party is clearly not over. Just digesting all the AI-generated slop or training AI to be better at producing readability and interpretability is going to take some work. And although I criticize these AI companies, new results do contribute to the advancement of the field. Not to mention that we have new directions in research and problems on education to figure out. I actually feel like there's more work to do now than before. If anything, it seems that we need more people in these fields.

People in academia should be discussing how to reform the system together, rather than spreading anxiety among ourselves.,There have also been plenty of discussions in the TCS community on how AI may affect the academic credit system and how we need to change. There is no final conclusion yet but I’m hopeful of a change soon.

And even if AI eventually becomes capable of doing almost everything, including writing and teaching, I still think we need people. The communities formed by human beings, and the synergy that comes from human interaction, cannot be replaced. I think those of us in STEM often underestimate the importance of people, just as people, not as instruments or thinking machines.

But none of this means that we don't need to take action. There will be many things that need to be done during this transition. The possibility of AI further exacerbating wealth inequality genuinely frightens me, as does the question of what will happen to primary, secondary, and higher education.

First, in the short term, I don't think people who already have jobs need to worry too much about losing them. In the long run, for a field like mathematics, where there are already so few positions to begin with, whether we end up with a few more or a few fewer jobs may not even be the most important question. It's not that I don't worry about the younger generation in our community. Rather,  for our field as a whole, what matters most may be finding a way for the higher education system to reform itself amid the current chaos and continue to exist.

Preserving this education system isn't just about making sure that we and the generations after us don't lose our jobs. It's about the fact that education and learning themselves need to be emphasized in the age of AI.

If society completely withdraws its support for learning simply because learning has lost its "instrumental value," and in doing so deprives people of the right to learn, that would be the truly terrifying outcome.

I believe that social activity is something a lot of us should consider learning about. 

One thing we really need in this era is more public intellectuals and more public discussion.

In other aspects, if people have the time, it would be worthwhile to learn more about unions and other collective bargaining organizations they could get involved in. We could also do more AI outreach and education for the general public, especially teenagers, helping people understand how to use AI effectively as a tool while preserving their ability to think independently.

We could also support organizations that defend digital privacy, such as the Electronic Frontier Foundation (EFF).

I really believe we need to have more imagination. There are ways of coexisting in this world beyond pure competition, and humanity has survived to this day in large part through cooperation. Viewing the world entirely through the lens of social Darwinism is incredibly narrow-minded.

We need to set an idealistic goal and work toward it, so that we can eventually converge to a not idealistic but reasonably good outcome. If right from the beginning,  we surrender to reality, or to the darker sides of our own nature, then we can only ever converge toward something worse.

=======================

To be honest with myself, I haven't spent much time thinking deeply about math problems lately. Maybe I've been using AI too much and my brain has gotten a bit rotten too.

 I also haven't been doing enough social work recently. Most time I’ve been just having anxiety and anger. I really need to get myself together.

In any case, I'm not going to join the AI rat race, and I'll also try my best not to spread anxiety to others. I will try to simply take actions from both an academic and social perspectives.

I believe there are many ways of us coexisting in this world that go far beyond zero-sum games and the law of the jungle.

Finally, I want to end with something a friend of mine once said (slightly modified):

My friends and I are on an island. We do our best to keep living here. If a storm comes, we'll do everything we can to resist it. But if in the end, the storm overwhelms us, then so be it.