Topic

17 July 2025

AI reflects back an image of ourselves, says Petr Koňas

In the spirit of our conversation, Petr Koňas provided a picture of his avatar. | Autor: DALL-e

In many fields, we are reaching the limits of human cognitive abilities. Artificial intelligence can help us process the enormous amount of information humanity has at its disposal. That’s the belief of Petr Koňas, an expert in numerical simulations and AI, who has recently taken on the role of AI ambassador at BUT. Thanks to his perspective, he is a moderate techno-optimist who hopes that the use of AI will primarily benefit humanity—whether through groundbreaking scientific discoveries or simply by making it easier to fill out a travel order.

Editor’s note: This interview is much longer than the usual texts on the BUT website. That’s because the topic is broad and the answers are so interesting that shortening it would be a shame. Please note in advance that it is a long read. If you're mainly interested in the AI ambassador role at BUT, skip to the last third.

While preparing for our interview, I asked the research tool Perplexity about you. What is your favorite tool that you use daily and that makes your life and work easier?

There are many tools, depending on what I’m currently focusing on. Perplexity is one of them. Then there’s LibreChat, which is a kind of aggregator that brings together most commercial and open-source language models, allowing me to switch between them as needed—from specialized models to more general ones. LibreChat also offers very useful so-called agents, tools designed to solve highly specialized tasks: for example, if I’m looking for a literature review, I summon the appropriate agent and my work becomes much easier. Recently, I’ve also been using RAGFlow for economic and administrative tasks at the Rectorate. My colleagues and I are now collaborating on how to apply AI in the area of accreditation.

And then there’s Python in general, because many tools are in a very raw state, and if I want to use their high-end capabilities, I have to customize them myself. Python truly plays an irreplaceable role in this.

Most of us became interested in AI only after trying ChatGPT for the first time, which was just a few years ago. But you’ve been working in the AI field for 30 years. What kind of AI was it back then, and how far has it come to enter everyday vocabulary today?

It’s a huge shift. My first encounters with AI were quite unremarkable—AI was just one of many statistical methods used for analysis, approximation, classification, and so on. For example, it helped me significantly in image analysis 30 years ago, because it could perform tasks like image segmentation much more efficiently. But back then, we lacked sufficient computational power. Even if the analysis brought qualitative insights, it took so long that you had to think twice about whether to let the computer work on it for a month or more. However, as hardware capabilities improved and computing power increased, so did the influence of AI—first in machine learning, then deep learning, and now large language models.

The use of language models combined with machine learning and all the other capabilities now available is absolutely irreplaceable and will play a tremendous role in the future. Because not only for scientists but for humanity as a whole, it offers a unique and fantastic opportunity: it can lead you from vague input to relevant data and help find solutions we either couldn’t discover before or had given up on because it took too long. I’m very glad to witness the current stage of AI and am excited about the unimaginable changes yet to come.

In many fields, we’re already hitting human limitations. We’re essentially unable to create synergistic knowledge that would truly advance us, let alone consolidate knowledge across disciplines...


You seem to be a techno-optimist, so I have to ask: do you think AI will bring about new, truly groundbreaking discoveries in fields where we’ve been waiting for them in vain for years—say, in astrophysics or particle physics? Is that one of AI’s potential roles?

Exactly. In many fields, we’re encountering human limits—not just cognitive, but also in terms of processing the sheer volume of information being generated. Just consider our own institution—how many articles and insights does BUT produce every year! No one person can process all that data in a way that evaluates the full extent of our knowledge. It’s always just a fragment. We’re essentially unable to create synergistic knowledge that truly pushes us forward, let alone consolidates insights across various disciplines. I fear the human brain is simply no longer up to the task—we need tools that can help. AI can do this; it can go deep and across disciplines, revealing connections and relationships that humans might never notice.

Recently, Tomáš Mikolov made an unfortunate comment about the “autism” of some scientists. But we must admit that, as researchers, we often reach a point where the general public has no chance of understanding not only the meaning but also the application of our results. In this regard, AI can once again play an irreplaceable role by making our knowledge accessible to the public.

I can confirm that—when I interview physics colleagues, I use AI to help me understand the technical texts we’re discussing. And it really does an excellent job of “translating” them into popular language.

You’re right, but that’s also where a great danger lies—namely, that we might remain “on the surface.” We need to think about how to move humans forward in the context of knowledge. And that’s a huge opportunity for us as an educational institution: how can we help our students master material within certain limits of time and quality? Here too, AI can help—it has the capacity to engage with each student individually, something a human teacher often can’t do.

The Faculty of Mechanical Engineering of the future according to AI. | Author: ChatGPT

Human Imperfection

I don’t want to dwell too much on the fears AI provokes—there are many. But let me ask just one: what do you think will remain uniquely human if we delegate many of our tasks and skills to AI?

That’s a very tough question. I think humans, like the rest of nature, are subject to evolution. And we’re now bumping up against many of our limitations because of the enormous leap humanity has made over the last 100 years. The shift has been so extreme that our physiological constraints are holding us back. I ask myself: how can we overcome this handicap? And I believe that AI is a natural next step forward.

There’s a risk that we’ll create a more advanced version of our own imperfections—and that’s not a comforting prospect. But as a scientist, I’m an optimist...


But you’re right—sometimes I wonder if development isn’t moving too fast, and whether we’re ready for it. The public’s reactions show that often, we’re not. We don’t realize that what AI reflects back to us is essentially a mirror of ourselves. What we see as errors are actually human flaws. We shouldn’t focus only on adapting AI, but also on exploring the root causes of our problems. If we don’t work on that, we risk creating a more advanced version of our own imperfections—which is certainly not a positive outlook. But as a scientist, I’m optimistic and believe that AI will help uncover the contradictions we have as a species and have been overlooking. Perhaps it will lead to a kind of catharsis—a realization of what it means to be human and what defines us as people.

You mentioned rapid development, so let me ask about the fear some experts have that AI might degenerate due to the increasing amount of generic content—i.e., that it could harm itself by training on its own outputs. And at the same time, humanity’s dependence on this technology is growing...

I understand the concerns of data analysts pointing this out. On the other hand, I’d add that even many scientific articles show signs of repetition and redundancy. And yet we’ve managed to use them as training data, as a baseline. I see a kind of internal contradiction here, but of course, the more redundant the data, the less AI will progress. It’s a natural cycle where quantity and quality alternate. Even in neural networks, we see that in some cases we need vast data, and in others, a breakthrough discovery changes how that data is processed or interpreted. That moves us forward.

Currently, we’re still talking about AI—not AGI (Artificial General Intelligence, which aims to replicate human cognitive abilities, editor’s note). When do you think AGI might be developed? Or will we continue using narrow AI tools for specific tasks?

Soon after language models became available, it became clear that their limitations lay in how they perceive the world. Text-based models only handle text. Then came speech processing, then image processing, and the number of modalities through which we describe the world keeps growing. The more they grow and interconnect, the more new relationships and new kinds of knowledge are uncovered. And the closer we get to a universal AI that can represent our perception of the world. If this condition is met, and if we also manage to develop a kind of shared motivation, I believe AGI will be created within a relatively foreseeable timeframe. But by that, I mean 20 years or more.

How AI imagines the BUT Rector's Office in 2050. | Author: ChatGPT

Everyone Deals with Travel Orders

Bringing things back down to earth: what does AI mean to you personally? And how did you come to relate to it?

To me, it’s a tremendous helper—it makes my work easier. What used to take months, then weeks, now takes minutes. I can do a literature review with a click. That leaves me time for things I’ve been putting off. I think this is a big challenge for the university and research in general—by removing bureaucracy and other non-creative activities, we can focus more on creativity, which I see as the defining trait of our humanity.

Let’s look at your position as AI coordinator at BUT. How did you come into that role—beyond your expertise, of course?

I stumbled into it almost by accident. (laughs) I was helping colleagues at the department, then it spread to the faculty, and suddenly Vice-Rector Vítězslav Máša contacted me—he’d heard about me and was working on AI strategy and development for the university. He has a very thoughtful and rational approach. I hope everything we discussed gets implemented. In any case, one thing led to another, and I took on the role.

What is your role supposed to be?

The main idea is coordination. Everyone today has their own perspective on AI and uses their own tools and methods. Unfortunately, this leads to duplication and wasted time on problems already solved. My role is to unify information, build a knowledge database on AI, and serve as a point of contact. We need to accelerate development in this area because, let’s be honest, Czech universities are a bit behind compared to the West. I see my primary task as meeting with people, forming teams focused on specific AI areas, and helping them transition to AI. The goal is to make AI a true helper, not a problem to overcome.

Is the creation of an AI assistant for BUT on the table—one you could ask how to fill out a travel order, for example?

When we first started talking about AI, I assumed the main concern would be its use in research. But I couldn’t have been more wrong! The biggest concern for everyone is administration. The travel order was mentioned by almost everyone. These are tasks that slow us down and would be great to automate. So yes, it’s a top priority. But I’m not sure there’s a simple, quick fix. Once you start digging into all the associated regulations, it’s overwhelming.

But you’re right—and the companies that reach out to me most often ask how to process and make large volumes of information and internal know-how accessible so that it doesn’t disappear when employees leave. AI is perfect for that—it can gather, organize, and lead you to your goal through a simple query.

What can BUT staff reach out to you for right now?

Definitely training requests—I already do those and will continue. But more generally: if you’re working with AI, get in touch. It doesn’t matter if the problem is big or small—even a minor issue can lead to something interesting. I’m happy to help anyone, as much as I can.

At the same time, I don’t claim to know everything happening in AI, even though I try to stay up-to-date. So I’d really appreciate sharing experiences with others as well.

You seem genuinely enthusiastic—what excites you about AI?

I’ve had the rare opportunity to witness several revolutions that changed humanity: the arrival of computers, the internet, and now AI. In my field of numerical simulations, development has been incredible. I enjoy observing these changes and staying in touch with people who are driven, creative, and full of new ideas. That’s always incredibly enriching. When you isolate yourself, it helps no one. But when you have the chance to communicate—even with the best—it’s something I recommend to everyone.

(ivu)

Source: Faculty of Mechanical Engineering BUT

Themes

Related articles:
Robots are meant to help us, not replace us, says Stanislav Věchet
Obelix flew to the United States
Auxetic material behaves differently than you would expect, says student Tomáš Vítek
Majáles returns after 2 years, royal candidates from BUT are ready to fight
A special tribometer customised by Brno scientists for a Japanese company