INTERVIEW
HELDER COELHO
Professor, researcher, and one of the pioneers of Artificial Intelligence in Portugal
By Bruna Ferreira, Storyteller & Content Manager at Quidgest
We are better than ChatGPT – we still are
This statement is from Helder Coelho, a professor, researcher, and one of the pioneers of Artificial Intelligence in Portugal. Despite a ‘forced retirement’ from teaching, he still maintains his office at the Faculty of Sciences at the University of Lisbon. It was there that we met him to discuss the surge in Generative AI, the future of human-machine symbiosis, the promises of Neuroscience, the ‘collateral damage’ of the technological revolution, the emergence of regulation, the crucial role of education, cinema as a reflection of reality, and the notion (or fantasy) of human extinction.
An interview that delves into the past but returns to the future with the same fluidity and enthusiasm of someone who shapes their life through critical thinking and interdisciplinary collaboration between AI and other fields of knowledge. Otherwise, “how could we possibly cope with the complexity of our world?”, the interviewee explains.
Where does your immense fascination with Artificial Intelligence (AI) come from?
I began my Engineering course at the Instituto Superior Técnico in 1961 and completed it in 1968, which means that instead of the usual six years, I took seven – and I also experienced Student’s Day. All of this gave me a certain direction. I was also actively involved in the Student Association, which allowed me to ‘customize’ the course curriculum, introducing new subjects and canceling others… despite some resistance from the professors (laughs). This was because the course was very general, covering both Electricity and Electronics, two very distinct areas. I chose Electronics, even though I faced challenges, like spending a year in the Netherlands and getting caught up in the war (Portuguese Colonial War).
In the midst of all this, my journey in AI began when I met Luís Moniz Pereira, around 1966. He was the one who pulled me into Cybernetics and then pushed me into AI, eventually becoming my mentor. After that, everything evolved naturally… I went to work at the National Laboratory of Civil Engineering (LNEC) and the first courses and the Computer Science department began to emerge.
At that time, personal computers were a rarity, and the Web was yet to be invented…
Yes, to give you an idea, the initial movements of AI emerged in 1955, but it was only officially founded in July 1956. This period was marked by a group of 10-15 notable figures, such as John McCarthy, Marvin Minsky, Herbert Simon, and Alan Newell. I got to know all of those I mentioned and even saw some of them in person when they came to Portugal for conferences.
My trip to Edinburgh (with Luís Moniz Pereira) was no coincidence; the University of Edinburgh was the first in Europe to focus on AI advancements and establish its own department. When we returned to Portugal, requisitioned by the LNEC, we contributed to the introduction of the new DEC-10 computers, identical to those used at the University of Edinburgh, and thus, much more advanced and ideal for stimulating the technological leap that LNEC would make in the field of Computer Science.
After over a decade at LNEC (1973-1989), having completed my PhD in AI in between (in 1980, under the guidance of David H. D. Warren at the University of Edinburgh), I realized that the institution had changed and Computer Science itself was a transforming field – much smaller and much more powerful computers. Instead of being a computer operator, I chose the teaching path. I had already taught in Luanda during the war, and then I chose the ISEG – Lisbon School of Economics and Management, where motivated by the intersection between Economics and AI, I taught for over 10 years. Subsequently, in 1995, I moved to the Faculty of Sciences in Lisbon.
You mentioned John McCarthy earlier, who stated: “AI is making a machine behave in ways that would be called intelligent if a human were so behaving”. Do you believe we have reached this point with chatGPT?
No. At present, the most advanced companies in AI (Microsoft, Apple, Amazon, IBM, etc.) are focused on selling services, tools, or computers. That’s their business. ChatGPT is another example: it originated from an Elon Musk company, which has grown and now boasts multi-million dollar investments from Microsoft. In 2022, there was a notable leap with the emergence of GPT-4 and chatGPT, which is slightly different. But this technology has been operational since 2020. Why didn’t they release it earlier? Because it wasn’t yet marketable.
So, of course, AI is now more accessible and usable, but many chatGPTs have also been sold. There has been progress, just not as much as they want us to believe. This interaction between humans and robots is not yet equivalent to the complete replacement of people by machines. We still have a long way to go before reaching a level of AI that can truly be called “intelligent” in the human sense. I even believe that the leap in 2014 was much more significant because it showcased the power of information processing and the increased speed of computers, as we saw in James Cameron’s “Avatar”. There were notable advancements in language translation by Google, allowing for more fluid and effective communication in different languages, and these foundations are crucial for a deeper understanding of language – a topic I explored in the past with my natural language processing program TUGA, which enabled translating conversations and interacting to provide an AI library service.
In summary: natural language processing is simpler, it’s more powerful, but it’s mostly statistics at work and little else. We are still, as Noam Chomsky believes, far from unraveling the true mystery behind intelligence, the power of human sensations and emotions, and the involvement of AI with Cognitive Science and Neuroscience.
In Portugal, AI research began at LNEC, led by you, along with professors Luís Moniz Pereira (whom you’ve already mentioned) and Fernando Pereira. It was the year 1973. Fifty years later, is AI still an exciting topic for those who study it?
Yes. On one hand, we see advancements in understanding that things are more complicated than they were before. We were also more naive, that’s true. But, and I fully agree with Chomsky, we haven’t reached the point yet. Natural language, speaking, is much more complex than we imagine. In other words, our lack of knowledge is much stronger than our understanding, and this might be one of the main reasons for such enthusiasm.
Right now, for instance, I’m very interested in consciousness. This is because I read a book by a South African neuroscientist (Mark Solms, “The Hidden Spring: A Journey to the Source of Consciousness”, Profile Books, 2021) who, using computer science, has made advancements in the topic of consciousness – yet, he’s still far from fully understanding the concept. Especially since even animals possess consciousness. I had a dog for almost 13 years, and to this day, I can’t understand how he managed to comprehend everything I said. Often, we find it amusing that dogs pretend and do certain things, but they do it knowingly.
You often mention that Apple products became bestsellers because they didn’t require computer knowledge to be used. Is this what happened with chatGPT or was it something else?
It was marketing. ChatGPT is stronger, and more capable, but not smarter. It’s just more comprehensive because it interacted with more data and learned more. It’s what happens to us: when we read books, we expand our knowledge and become more capable. ChatGPT read many books. It knows more about many topics than we possibly do. And then? It’s still not intelligent. It seems to be, but we are better than chatGPT – we’re still better. So, do we have more information? We already know we do. Can we use this information? We can. Are we sure it always gets it right? No. Does it fail? Yes, but we also make mistakes from time to time. So, between us and the chat, we might still have more confidence in ourselves.
When we move from intelligence to consciousness, we enter different territory because we learn not just with our eyes, in the case of reading, but also with our sensations. Vision is fundamental, but how it supports our intelligence, we still don’t know. There are advancements, but not as significant as we imagine. I’ll use the mashed potato metaphor: you still need to stir a lot with the spoon to get to a stronger and smarter product.
And here I move to another area: complexity. When I visited the Santa Fe Institute in 1997, I realized how complexity is an extremely interesting challenge. It still is! In all areas. In politics, for example, the most interesting thing is complexity – seeing how politicians clash with each other, pull each other, etc. The problems of our world are complicated, they are complex – like climate change, for instance, and few seem to be bothered by it because they think they’ll die in some time and the problem will be for the next generations.
From the beginning, humans sought ways to communicate with machines – first with wires, then through low, high, and very high-level programming languages (like Prolog, which you explored in your Ph.D. and in a book), with less and less code. Today, it’s even possible to create digital solutions without knowing code, through platforms like Quidgest’s Genio. Is the future of this man-machine symbiosis in the elimination of barriers and layers between the two?
I’m convinced of that. Robots have evolved significantly, although I’ve never been very fond of robots because I much prefer programming than watching a robot’s antics. But the truth is that all these machines were not created to replace humans, but to assist us. Humans are not all the same – some are content with science, discovering new things; others take advantage of what the former do and exploit it to profit and live, for a while, rich and happy. Others still prefer to read and think about the subject.
I’m much more interested in thinking about things, trying to discover other things, and being entertained. Because when we work in Science, that’s what we do: we entertain ourselves. I finished my professorship in 2014 and I’m still here in my office. I really enjoy continuing to come to the faculty and they haven’t locked me out yet (laughs). I’m privileged and my colleagues are very kind to me. For my part, I would also continue to teach… Much more than I currently do. Hence, I now spend more time writing texts so that, perhaps, someone might read me.
Neurosciences are another passion of yours. To what extent is it possible for machines to understand and be part of the human brain today, and what kind of advancements can we expect?
At the moment, what we can say is that we still don’t know how to access the brain or how to remove the brain and place it elsewhere. However, from the perspective of interfaces, we have witnessed significant progress. Miguel Castelo-Branco, a physician and neuroscientist at the University of Coimbra, has been doing extraordinary work in the field of autism, focusing on concepts of neurodiversity and personalized medicine, for example.
It’s interesting because we realize that there’s a convergence between scientists and those who need support. This puts Neurosciences at the forefront, as there’s still so much to discover and accomplish. Science shouldn’t only be used for harm or to take lives; there should be a focus on harnessing it for the greater good.
Following what you just mentioned, do you fear that AI might have a devastating impact on employability, or on the contrary, do you believe it can help address the talent shortage?
AI emerged due to the curiosity of a few individuals (that list of 10, 15 people I mentioned earlier) who wanted to know more because they felt Cybernetics wasn’t enough, and they tried to integrate more disciplines. This remains intriguing even today because if we look at universities, very few understand the advantage of interdisciplinarity. Fortunately, at the University of Lisbon, we have Master’s and Doctorate programs in Cognitive Science that combine Neurosciences, Philosophy, Psychology, Computer Science, and AI. What does this mix of disciplines allow? It allows us to think. Because when disciplines intersect and complement each other, they make us think. And in education, students sometimes don’t think enough.
I taught for 43 years. I encountered exceptional students, the kind every teacher would love to have, and others who weren’t interested in anything. This continues to happen today, without much intervention from the broader circle of deans, who focus on matters that don’t genuinely impact what a University is about – which is what? It’s not about getting grades to finish a course. It’s about preparing individuals who can think and navigate the complexity of our times.
“The artificial and the natural have always been subjects of philosophical discussion” is a statement of yours. As the artificial gains more visibility, quality, precision, and utility across various domains, what remains for humans and natural intelligence?
An infinite array of things remains. Notice that, at this moment, we’ve returned to space exploration, despite all the complexities present on the planet we inhabit. That is, challenges and problems repeat over time (and some even tend to worsen), which compels us, as humanity, to reflect on what we need to do and what we haven’t done yet. Meanwhile, situations deteriorate, whether it’s a road flooded with water or a forest fire. Technology alone doesn’t work miracles.
Fake news, information bias, hallucinations, copyright issues, security, or hacking are some examples of the ‘collateral damage’ inherent in technological evolution. Regulation seems to be the only way forward, but how can we achieve global uniformity, considering the social, political, and economic differences between Europe, the USA, and China?
It’s true that there are rifts between these three blocs. But there are reasons for these rifts. For instance, let’s consider China and the USA. For a while, the Chinese sold everything imaginable to the Americans and allowed Chinese students to study in the USA. Not anymore. Why? Because even though China hasn’t yet overtaken the USA competitively, it’s close. They now publish articles in the most recognized venues, and in greater numbers, than the Americans. At the company level, I’ve made a list, and I still believe the USA has the strongest companies. But this is changing because the Chinese are now capable of doing the same as the Americans… They learn and innovate very quickly. And there are many more of them.
Can we talk about pre-algorithmic and post-algorithmic critical thinking? What does your experience as a professor tell you, having always promoted the idea of “building bridges, opening paths, digging trenches, and thinking ‘outside-the-box'”?
Thinking about complexity is always beneficial. It’s a pity that there are so few students concerned about this. There should be more. In 2001, I started a series of workshops in Arrábida (“Arrábida Workshop on Complexity”), with the latest edition taking place last July. I’m uncertain if we can continue, as we need to secure two or three thousand euros in funding, but it has been a very fruitful initiative. The participants are generally professors, and the topic of discussion is complexity. It’s an immersive experience because we go to a place with only those individuals, and we always achieve a very positive outcome.
In our last meeting, we discussed the different departments of the Faculty – the connections, the separations, and what needed improvement. We also looked at the Biology departments, the ongoing projects, and the challenges of securing foreign funding to advance these projects. You can’t imagine how complicated this is.
What we can say is that we are far from having a university capable of engaging with the country and making a significant impact, for reasons we all understand. Our universities, even those considered good abroad in the rankings, rank quite low… They are good, but not as “good” as others that are “better”. The Italians have a rule: those who earn a doctorate at an institute cannot teach at that institute. They must go elsewhere. I bring up this example because a university also becomes good when it has professors, researchers, and students from other universities. The “homegrown talent” isn’t always the best. We return to the same issue of intersecting areas and different backgrounds, experiences, institutions, etc.
Still on education, you believe that “it’s necessary to generalize and trivialize computer science education” and that “a country’s ability to manipulate (understand, process, apply, transfer, and transmit) information largely determines its future”. What are we doing in Portugal to create a promising future in this regard?
We’re not doing enough. Just the other day, I was discussing with my colleagues that the only way to achieve this is by promoting diversity, not excluding institutions like Católica, Porto, Coimbra, or other institutions and professors. In other words, by bringing together people and ideas that enrich and add to the discussions. And this can only be achieved when we have diverse opinions.
Taking advantage of your cinephile side, I ask: since “Metropolis” (1927), the first feature film to introduce a robot, which movie would you recommend to those interested in AI?
“Oppenheimer” by Christopher Nolan. Because the atomic bomb is a paradigmatic case, and this film follows, from start to finish, the physicist, the scientist behind its origin. It’s powerful in the sense that it shows how he reflects on what he has to do. This kind of thinking is about complexity, objectives, consequences, about the value of human life… And all these issues are very well seen and explored by Nolan, who for me is one of the great directors of our time.
While James Cameron’s “Avatar 2”, another film I loved, is more impressive from a technological standpoint, “Oppenheimer” focuses on the character and his internal conflicts. It’s also a film that challenges us to think about the historical episodes of destruction that repeat themselves over and over again, for very similar reasons and ambitions. Think about the invasion of Ukraine. It’s unbelievable. Using cinematic language, it makes us say: “I’ve seen this movie before.”
*This interview was originally published in the Mais Superior magazine.