Knowledge as a Personal Commitment

10 Aug 2026

It’s pretty clearly the last year of math and theoretical computer science research in the style we’ve known it.

— Scott Aaronson

Personal knowledge is an intellectual commitment, and as such inherently hazardous [...]. Into every act of knowing there enters a passionate contribution of the person knowing what is being known, and this coefficient is no mere imperfection but a vital component of his knowledge.

— Michael Polanyi

I am both fascinated and anxious to see how recent progress changes the academic scene, especially as I find myself in a relatively unique position where 3.4 years of my PhD struggle so far are almost perfectly aligned with AI development. Now, when I look at my open career, I feel some level of uncertainty combined with excitement, and I have serious questions about my role (as a human being) in the development of science.

I remember my excitement about Grammarly being able to correct my typos while writing my master’s thesis, and my inner voice announcing that my end-times theory was that this was the most computers could assist with the conduct of research. Yikes.

Then, I remember grading students’ assignments and some frustration about students submitting AI slop. I was relatively convinced back then that making the exercise relatively obfuscated by using non-standard notation and making sure that every detail of the solution is present (not hand-waved) were enough to distinguish human vs AI work. Yikes.

Recently, I don’t remember where, I heard someone complaining that the 3SAT to SVP reduction that the OpenAI model produced is basically a mixture of known tricks, and insisting that human creativity is still strictly more powerful than that of any reasoning model. This time I don’t buy it.

My first moral thoughts about AI were, of course, related to social inequalities, concentration of power in a few stakeholders, ability to micro-adjust political (in a very broad sense) narrative and climate justice. Those are still extremely valid questions, yet I don’t focus on them as I perceive them as (pardon my oversimplification) relatively transitional. I.e. the same could have been said about many technological advances, and I don’t find AI’s role extremely unique.

My recent thoughts are more related to what the role of researchers (and probably also many industrial experts, like engineers, architects, planners) is if, essentially, AI is (or soon will be) better in nearly every aspect. I have no doubts about it. We can only slow it down.

Doing research has a lot in common with discovering human fragility, i.e. what we don’t know or understand. Often, I can see some phenomenon in math, but I cannot explain it. I can follow some steps, but cannot get intuition or convince anyone else. And there are many similar occurrences, which can be summarised in a simple observation that our resources are limited, i.e. we have limited time, energy, intellectual capabilities, and attention spans. Paradoxically, I believe this is the strength of humanity, because it allows us to use those limited resources as a form of currency in a society. And because that currency is limited, it’s extremely valuable.

I try to ask myself what my role is as a researcher. I don’t believe that my main duty is to produce correct research. I would like to shape the world, or more precisely, try to impact the world in a direction which is, I believe, the right direction for humanity. Regardless of how pompous it sounds, I believe that this is also the case for my research. I work on crypto not (only) because those are fun math/eng problems, but because I believe that we, as humanity, need this research. Not necessarily mine, but cryptographic research as a whole, to live in a good world.

In other words, I put into cryptography what is most valuable for me – time, energy, attention, care – and this is so valuable because I know I could have done so many things with my time otherwise. I choose to pursue my goal, while declining to do many things that would also contribute to humanity. E.g., maybe I should spend my life as a social worker? I answer negatively because I believe that this is the most valuable usage of my energy.

I am a human with limited time, and I can understand very little. I spent enormous effort to understand something very small (in terms of global knowledge), and this effort is the price. I design some protocols, implement them and put my credibility on them, being aware that this does not say anything more than “I spent many years on that, I made many sacrifices, I could have done something else, but I didn’t, and I believe it’s worth sharing”. This is the stamp that AI cannot affix, and I believe this is also what makes scientific credibility not merely a proof of correctness, but rather a social concept.

More concretely, I am responsible for my research output. I create something new, and I certify (with my reputation, experience and credibility) that this is an outcome I am happy to put my name on. If broken, it’s on me. If the majority of my output is unreliable, I won’t get a good job. If I don’t get a good job, my family may starve. If I mess up hard, one can bring me to court, and I go to jail. If my research is stable, I may become famous and enjoy some level of prestige.

So does it really matter what entity – human, AI or whatever – produces the result? More important is how the results get trusted. What makes them believable? What turns logical puzzles into a new generation of protocols? Who are those that we trust?

It seems that the production of scientific results might be offloaded outside society, but I am relatively confident that the expertise, understood as societal trust and the ability to pay with societal currency (trust, time, effort, attention or reputation), is an inherently human feature.

The last point to emphasise is that I understand expertise very broadly. It is not only the ability to verify technical correctness or novelty (again, AI is/will be better). Rather, it’s an ability to bring the research topic into a community, to gather fellows around some idea, and convince them of the relevance of some research. To opine (solely or jointly) about methodology, ethics, rigour or societal value of the contribution. Even further, to express the willingness to exist in a world where some ideas are implemented. Those are skills which can be broadly understood as turning knowledge into society, which only humans, as only members of society, can do.