Synthetic intelligence is on all people’s lips as of late, sparking pleasure, concern and limitless debates. Is it a drive for good or unhealthy – or a drive we even have but to completely perceive? We sat down with outstanding laptop scientist and AI researcher Mária Bieliková to debate these and different urgent points surrounding AI, its impression on humanity, and broader moral dilemmas and questions of belief it raises.
Congratulations on changing into the newest laureate of the ESET Science Award. How does it really feel to win the award?
I really feel immense gratitude and happiness. Receiving the award from Emmanuelle Charpentier herself was an unimaginable expertise, full of intense feelings. This award would not simply belong to me – it belongs to all of the exceptional individuals who accompanied me on this journey. I imagine they had been all equally thrilled. In IT and likewise in applied sciences usually, outcomes are achieved by groups, not people.
I am delighted that that is the primary time the primary class of the award has gone to the sector of IT and AI. 2024 was additionally the primary 12 months the Nobel Prize was awarded for progress in AI. In actual fact, there have been 4 Nobel Prizes for AI-related innovations – two in Physics for Machine Studying of Neural Networks and two in Chemistry for coaching Deep Neural Networks that predict protein constructions.
And naturally, I really feel immense delight for the Kempelen Institute of Clever Applied sciences, which was established 4 years in the past and now holds a secure place within the AI ecosystem of Central Europe.
A number one Slovak laptop scientist, Mária Bieliková has performed intensive analysis in human-computer interplay evaluation, consumer modelling and personalization. Her work additionally extends to the info evaluation and modelling of delinquent habits on the internet, and she or he’s a outstanding voice within the public discourse about reliable AI, the unfold of disinformation, and the way AI can be utilized to fight the problem. She additionally co-founded and at present heads up the Kempelen Institute of Clever Applied sciences (KInIT), the place ESET acts as a mentor and partner. Ms. Bieliková lately received the Excellent Scientist in Slovakia class of the ESET Science Award.
Writer and historian Yuval Noah Harari has made the pithy observation that for the primary time in human historical past, nobody is aware of what the world will appear like in 20 years or what to show in colleges at present. As somebody deeply concerned in AI analysis, how do you envision the world twenty years from now, significantly by way of expertise and AI? What are the talents and competencies that can as soon as be important for at present’s kids?
The world has at all times been tough, unsure, and ambiguous. At present, expertise accelerates these challenges in ways in which folks wrestle to handle in actual time, making it onerous to foresee the results. AI not solely helps us automate our actions and exchange people in numerous fields, but in addition create new constructions and artificial organisms, which might probably trigger new pandemics.
Even when we didn’t anticipate such situations, expertise is consciously or unconsciously used to divide teams and societies. It is now not simply digital viruses aiming to paralyze infrastructure or acquire assets; it is a direct manipulation of human considering by way of propaganda unfold on the velocity of sunshine and magnitude we could not have imagined a number of a long time in the past.
I don’t know what sort of society we’ll stay in 20 years from now or how the principles of humanity will change. It’d take longer, however we would even be capable of alter our meritocratic system, at present primarily based on the analysis of information, in a means that doesn’t divide society. Maybe we’ll change the best way we deal with information as soon as we understand we won’t absolutely belief our senses.
I’m satisfied that even our youngsters will more and more deviate from the necessity for data and evaluating success in numerous checks, together with IQ checks. Data will stay necessary, however it should be data that we will apply. What is going to really matter is the vitality individuals are keen to put money into doing significant issues. That is true at present, however we frequently underutilize this angle when discussing schooling. We nonetheless consider cognitive expertise and data regardless of realizing these competencies alone are inadequate in the actual world at present.
I imagine that as expertise advances, our want for robust communities and the event of social and emotional expertise will solely develop.
As AI continues to advance, it challenges long-standing philosophical concepts about what it means to be human. Do you assume René Descartes’ commentary about human exceptionalism, “I believe, due to this fact I’m”, will should be re-evaluated in an period the place machines can “assume”? How far do you imagine we’re from AI methods which may push us to redefine human consciousness and intelligence?
AI methods, particularly the massive basis fashions, are revolutionizing the best way AI is utilized in society. They’re frequently bettering. Earlier than the tip of 2024, OpenAI introduced new fashions, O3 and O3mini, which achieved vital developments in all checks, together with the ARC-AGI benchmark that measures AI’s effectivity in buying expertise for unknown duties.
From this, one may assume that we’re near reaching Synthetic Common Intelligence (AGI). Personally, I imagine we aren’t fairly there with present expertise. We’ve got wonderful methods that may help in programming sure duties, reply quite a few questions, and in lots of checks, they carry out higher than people. Nevertheless, they don’t really perceive what they’re doing. Due to this fact, we can’t but speak about real considering, regardless that some reasoning behind process decision is already being achieved by machines.
Simply as we perceive phrases like intelligence and consciousness at present, we will say that AI possesses a sure degree of intelligence – that means it has the power to unravel complicated issues. Nevertheless, as of now, it lacks consciousness. Primarily based on the way it capabilities, AI doesn’t have the potential to really feel and use feelings within the duties it’s given. Whether or not this may ever change, or if our understanding of those ideas will evolve, is tough to foretell.
The notion that “to create is human” is being more and more questioned as AI methods turn into able to producing artwork, music, and literature. In your view, how does the rise of generative AI impression the human expertise of creativity? Does it improve or diminish our sense of id and uniqueness as creators?
At present, we witness many debates on creativity and AI. Individuals devise numerous checks to showcase how far AI has come and the place these AI methods or fashions surpass human capabilities. AI can generate photos, music, and literature, a few of which could possibly be thought-about artistic, however definitely not in the identical means as human creativity.
AI methods can and do create unique artifacts. Though they generate them from pre-existing supplies, we might nonetheless discover some really new creations. However that is not the one necessary side. Why do folks create artwork, and why do folks watch, learn, and hearken to artwork? At its essence, artwork helps folks discover and strengthen relationships with each other.
Artwork is an inseparable a part of our lives; with out it, our society can be very totally different. That is why we will admire AI-generated music or work – AI was created by people. Nevertheless, I don’t imagine AI-generated artwork would fulfill us long-term to the identical extent as actual artwork created by people, or by people with the help of expertise.
Simply as we develop applied sciences, we additionally search causes to stay and to stay meaningfully. We would stay in a meritocracy the place we attempt to measure the whole lot, however what brings us nearer collectively and characterizes us are tales. Sure, we might generate these too, however I’m speaking in regards to the tales that we stay.
AI analysis has seen fluctuations in progress over the a long time, however the latest tempo of development – particularly in machine studying and generative AI – has surprised even many experts. How briskly is simply too quick? Do you assume this speedy progress is sustainable and even fascinating? Ought to we decelerate AI innovation to higher perceive its societal impacts, or does slowing down threat stifling useful breakthroughs?
The velocity at which new fashions are rising and bettering is unprecedented. That is largely because of the means our world capabilities at present – an enormous focus of wealth in non-public corporations and sure elements of the world, in addition to a world race in a number of fields. AI is a big a part of these races.
To some extent, progress is dependent upon the exhaustion of at present’s expertise and the event of latest approaches. How a lot can we enhance present fashions with recognized strategies? To what extent will large corporations share new approaches? Given the excessive price of coaching giant fashions, will we simply be observers of bettering black bins?
At current, there isn’t any stability between the methods humanity can create and our understanding of their results on our lives. Slowing down, given how our society works, is just not potential, for my part, with no paradigm shift.
That is why it’s essential to allocate assets and vitality to analysis the results of those methods and to check the fashions themselves, not simply by way of standardized checks as their creators do. For instance, on the Kempelen Institute, we analysis the skills and willingness of models to generate disinformation. Not too long ago, we have now additionally been looking into the generation of personalized disinformation.
There’s a variety of pleasure round AI’s potential to unravel international challenges – from healthcare to local weather change. The place do you imagine the promise of AI is biggest by way of sensible and moral purposes? Can AI be the “technological fix” for a few of humanity’s most urgent points, or can we threat overestimating its capabilities?
AI might help us deal with essentially the most urgent points whereas concurrently creating new ones. The world is filled with paradoxes, and with AI, we see this at each flip. AI has been useful in numerous fields. Healthcare is one such space the place, with out AI, some progress – for instance, in creating new drugs – wouldn’t be potential, or we must wait for much longer. AlphaFold, which predicts the construction of proteins, has monumental potential and has been used for years now.
However, AI additionally permits the creation of artificial organisms, which could be useful but in addition pose dangers corresponding to pandemics or different unexpected conditions.
AI assists in spreading disinformation and manipulating folks’s ideas on points like local weather change, whereas on the identical time, it could assist folks perceive that local weather change is actual. AI fashions can display the potential penalties for our planet if we proceed on our present path. That is essential, as folks are likely to focus solely on short-term challenges and infrequently underestimate the seriousness of the state of affairs except it immediately impacts them.
Nevertheless, AI can solely assist us to the extent that we, as people, permit it to. That is the most important problem. Since AI would not perceive what it produces, it has no intentions. However folks do.
With nice potential additionally come vital dangers. Outstanding figures in tech and AI have expressed concerns about AI changing into an existential risk to humanity. How do you assume we will stability accountable AI improvement with the necessity to push boundaries, all whereas avoiding alarmism?
As I discussed earlier than, the paradoxes we witness with AI are immense, elevating questions for which we have now no solutions. They pose vital dangers. It is fascinating to discover the probabilities and bounds of expertise, however then again, we aren’t prepared – as people, nor as a society – for such a automation of our expertise.
We have to make investments at the very least as a lot in researching the technological impression on folks, their considering, and their functioning as we do within the applied sciences themselves. We’d like multidisciplinary groups to collectively discover the probabilities of expertise and their impression on humanity.
It is as if we had been making a product with out caring in regards to the worth it brings to the patron, who can purchase it, and why. If we didn’t have a vendor, we would not promote a lot. The state of affairs with AI is extra critical, although. We’ve got use circumstances, merchandise, and individuals who need them, however as a society, we don’t absolutely perceive what’s occurring after we use them. And maybe most individuals do not even wish to know.
In at present’s international world, we can’t cease progress, nor can we gradual it down. It solely slows after we are saturated with outcomes and discover it onerous to enhance, or after we run out of assets, as coaching giant AI fashions may be very costly. That’s the reason their finest safety is researching their impression from the start of their improvement and creating boundaries for his or her use. Everyone knows that it’s prohibited to drink alcohol earlier than the age of 18, or 21 in some international locations, but usually with out hesitation, we permit kids to speak with AI methods, which they’ll simply liken to people and belief implicitly with out understanding the content material.
Belief in AI is a serious matter globally, with attitudes towards AI methods varying widely between cultures and areas. How can the AI analysis group assist foster belief in AI applied sciences and be certain that they’re seen as useful and reliable throughout numerous societies?
As I used to be saying, multidisciplinary analysis is crucial not just for discovering new prospects and bettering AI applied sciences but in addition for evaluating their expertise, how we understand them, and their impression on people and society.
The rise of deep neural networks is altering the scientific strategies of AI and IT. We’ve got synthetic methods the place the core rules are recognized, however by way of scaling, they’ll develop expertise that we can’t at all times clarify. As scientists and engineers, we devise methods to make sure the mandatory accuracy in particular conditions by combining numerous processes. Nevertheless, there’s nonetheless a lot we do not perceive, and we can’t absolutely consider the properties of those fashions.
Such analysis doesn’t produce direct worth, which makes it difficult to garner voluntary help from the non-public sector on a bigger scale. That is the place the non-public and public sectors can collaborate for the way forward for all of us.
AI regulation has struggled to maintain up with the sector’s speedy developments, and but, as somebody who advocates for AI ethics and transparency, you’ve possible thought-about the position of regulation in shaping the longer term. How do you see AI researchers contributing to insurance policies and rules that guarantee the moral and accountable improvement of AI methods? Ought to they play a extra energetic position in policymaking?
Fascinated by ethics in analysis is essential, not solely in analysis but in addition within the improvement of merchandise. Nevertheless, it may be fairly costly as a result of it will be important that an actual want arises on the degree of important lots. We nonetheless have to contemplate the dilemma of latest data acquisition versus the potential interference with the autonomy or privateness of people.
I’m satisfied {that a} good decision is feasible. The query of ethics and credibility should be an integral a part of the event of any product or analysis from the start. On the Kempelen Institute, we have now experts on ethics and regulations who assist not solely researchers but in addition corporations in evaluating the dangers linked to the ethics and credibility of their merchandise.
We see that each one of us have gotten extra delicate. Philosophers and legal professionals take into consideration the applied sciences and provide options that don’t remove the dangers, whereas scientists and engineers are asking themselves questions they hadn’t thought-about earlier than.
Basically, there are nonetheless too few of those actions. Our society evaluates outcomes based totally on the variety of scientific papers produced, leaving little room for coverage advocacy. This makes it much more important to create house for it. In recent times, in sure circles, corresponding to pure language processing or recommender system communities, it has turn into customary for scientific papers to incorporate opinions on ethics as a part of the evaluation course of.
As AI researchers work towards innovation, they’re usually confronted with moral dilemmas. Have you ever encountered challenges in balancing the moral imperatives of AI improvement with the necessity for scientific progress? How do you navigate these tensions, significantly in your work on personalised AI methods and information privateness?
On the Kempelen Institute, it has been useful to have philosophers and legal professionals concerned from the very starting, serving to us navigate these dilemmas. We’ve got an ethics board, and variety of opinions is one in every of our core values.
For sure, it’s not straightforward. I significantly discover it problematic after we wish to translate analysis outcomes into follow and encounter points with the info the mannequin was educated on. On this regard, it’s essential to make sure transparency from the outset, so we can’t solely write a scientific paper but in addition assist corporations innovate their merchandise.
Given your collaboration with giant expertise corporations and organizations, corresponding to ESET, how necessary do you assume it’s for these corporations to guide by instance in selling moral AI, inclusivity, and sustainability? What position do you assume firms ought to play in shaping a future the place AI is aligned with societal values?
The Kempelen Institute was established primarily based on the collaboration of people with robust educational backgrounds and visionaries from a number of giant and medium-sized corporations. The concept is that shaping a future the place AI aligns with societal values can’t be realized by only one group. We’ve got to attach and search synergies wherever potential.
For that purpose, in 2024, we organized the primary version of the AI Awards, targeted on Reliable AI. This occasion culminated on the Forbes Enterprise Fest, the place we introduced the laureate of the award – AI:Dental, a startup. In 2025 we’re efficiently persevering with the AI Awards and have obtained extra and better high quality purposes.
We began discussing the subject of AI and disinformation nearly 10 years in the past. Again then, it was extra educational, however even then, we witnessed some malicious disinformation, particularly associated to human well being. We had no concept of the immense affect this matter would ultimately have on the world. And it is solely one in every of many urgent points.
I concern that the general public sector alone has no likelihood of tackling these points with out the assistance of huge corporations, particularly at present when AI is being utilized by politicians to realize recognition. I think about the subject of trustworthiness in expertise, significantly AI, to be as necessary as different key subjects in CSR. Supporting analysis on the options of AI fashions and their impression on people is prime for sustainable progress and high quality life.
Thanks to your time!
