Interview to Professor Alessandro Verri
The following is my conversation with Prof. Alessandro Verri.
G: Why did you choose physics at the university?
A: My desire was to study medicine. I very much liked physics as well as mathematics. But during high school my idea was to become a doctor. Physics classes used to start one month earlier compared to other courses. I attended the first lessons, made some friendships, enjoyed it and got through.
G: How did you live your university years? Try to find three adjectives to describe them.
A: Certainly they have been interesting to me. It was the first time in my life when I enjoyed studying. Somehow it was also the first moment in which I studied in a proper sense, not with the objective of answering questions or getting a good mark (so that my mum could tell me “Nooo, you didn’t get an 8!” [laughs]). Thereby, I studied alone and with much pleasure. There have been some topics which I studied and enjoyed a lot. My university years have also been passionate. However, I still remember that I used to suffer very much at the moment of the exam, which to me was a moment of pain, stress. I definitely felt better when I was done.
G: Then, we can say your university years have been interesting, passionate, with the exceptions of the moments of exam.
A: Yes, excluding the moments of exam when I used to suffer, in the sense that I couldn’t sleep, I was worried, not calm. I would have regretted a failure! I pondered them with a lot of meaning. Now this makes me smile, but at that time…
G: How much weight did your study have with respect to your life outside the university?
A: [smiles] It was important! I prefer not to comment on other things I would have liked to do but didn’t manage to. Also because those are the typical things of a person aged twenty [laughs]. We can say that outside the university I had not the same success. On the other hand, I did pretty well in studying.
G: Now a question of projection on the present. If you had to enroll to the university today, would you choose computer science or again physics, or rather another thing?
A: That’s a weird question I have to admit, because I was back to 18 years old perhaps I wouldn’t know.
G: What about if it was an advice?
A: I wouldn’t suggest anything in particular. I believe that it’s not true that anyone should necessary follow what they have inside, however there are certainly people who have at least very clear intentions. And in that case I wouldn’t oppose to any of the possible choice.
G: And therefore projecting on you? Would you attend again physics lessons? Or perhaps would you snoop around computer science lessons?
A: I’ve always been more attracted by “hard” sciences. Likely, I would go back to physics. Computer science has its own “hardness”. It would certainly be in the crosshairs. Certainly, with a different inclination compared to that of just programming. There have been topics which I really enjoyed along the years!
G: Let’s speak now about artificial intelligence. Who has been for you the luminary of artificial intelligence? A person who gave you special inspiration or whose studies have been really important to you. Maybe even a couple of people.
A: The person who has had the most influence on my professional life has certainly been Tommy Poggio. For the chats I’ve been fortunate to have with him. Not only for the collaborations and the things that we have done together. Having lunch with Tommy is invaluable! Because he has been a person who was able to transfer to me a perspective on things, a way of thinking, a way of being close to things which was very important to me! I owe Tommy the open-mindness, the capacity of not stopping, of seeing where things go, of living, in a sense, in the world out there. And not to forget that if you don’t face it nobody will acknowledge you. I also owe him the ability of sniffing where things go. He started dealing with learning when most people reckoned it was wasteful of time. I would like to add to the list, but in a different form, my advisor Vincent Torre, who taught me something that I have carried with me throughout my life. That is, to bite something when you have the feeling that it’s interesting and not to let go of the bone, to take out all the meat around it and then keep cleaning it up… And somehow I think this fact, as a Genoese person (i.e., a person who has spent his life in a place that is not MIT, Stanford), has allowed to carve out a little space. So we at MaLGa are, as to speak, a point on the map, and in relation to that, I think I have partly contributed to us becoming that. This would not have happened if I had not followed this style, which has never been an official advice, but has been a way of looking at one’s scientific life and of approaching one’s work.
G: And now we’re coming back to that because we want to continue talking about artificial intelligence, starting with a few ideas. The first question (which was also asked to Tomaso Poggio in his interview [*]) is: what is the goal you think we can achieve in creating intelligent systems? Will we be able to make progress without understanding how the brain works?
A: What I am saying is probably a bit disappointing because I would answer yes and no. What we see at the moment, which is rightly or mistakenly called artificial intelligence, answers the first part of the question. That is, you can do a lot of things without having the faintest idea of how the brain works. And one way of looking at this is that intelligence is overrated. Perhaps perception has been thought of as a problem or as a capacity intimately connected to the complexity of the brain and perhaps this is not true, it is much more reactive. And the ability to build responsive systems that work better than humans seems to me to be there for all to see… and it may even be possible to go much further! I have hesitations, on the other hand, about the fact that moving beyond perception and into the deeper layers of intelligence (e.g. emotions) it may be that we need a deeper understanding of how the brain works. You have to put a ‘may be’ in it because if they had interviewed me thirty years ago I would have answered - wrongly - that perception is a problem that could never be tackled without knowing the brain.
G: So, something you’ve come to your senses on?
A: Well, yes, the easiest way to see how dated the articles on which my scientific life has been built is that in them computer vision was a problem intimately connected to the fine and complex structure of the brain. I am not saying that this is not true, because it may be that there are still parts of vision, particularly the understanding of images, that require much more ‘depth’. But, on the other hand, it is also equally clear that many of the things that used to be said have collapsed with a fast computer and two million parameters… Not to make it trivial, but let’s say that it was essentially a technological advance, not a scientific one, that brought neural networks to where they are today. Those same networks were conceived a bit simpler, but not too much simpler, eighty years ago.
G: Let’s continue talking about artificial intelligence with two more insights, which are a bit more in the past and which I would like to share. Let’s try to project them into the present and the future. One is Marvin Minsky’s book, The Society of the Mind, in which he sees the mind as a society, i.e., a set of functions, of organs, of independent agents. The other cue is David Marr’s so-called ‘levels of understanding’, i.e. the three fundamental building blocks for approaching intelligence (computational, algorithmic and hardware). I would like to understand whether these visions anchored in the past perhaps have a maturity and even a future. In your opinion, can they still be considered ‘compasses’?
A: I would say that the first answer I gave [intelligence is overrated, ed] would certainly include these two contributions as, in some sense, dated. It’s unfortunate that that’s the case… [laughs], but they are dated because computing power was paltry in those years compared to what it is now, so certain things couldn’t happen. It’s always easy to dismiss content with hindsight, but I think Marr’s thinking was very influential for me. I liked it a lot! What has remained with me from that way of thinking, and which at this moment in time leads me to be in a resounding minority on the world stage (no one is shouting at me… [laughs]), is that the idea of the brain as a machine that has had millions of years to evolve resides in mathematical models that are a bit more sophisticated than those that are currently being used. I say that I am in the minority because right now it seems to me that the wave of artificial intelligence that works is to brute-force the figure that has made it possible to build systems that work well. What these current systems fail to emulate is what I call ‘perceptive intelligence’, linked to the ‘first tenth of a second’, which is capable of doing extraordinary things, in which one has a somewhat deeper understanding of reality. Let’s think for example of what is happening between us right now. I think that ‘perceptive intelligence’ is not only related to the ability to read how empathic we are or what kind of movements we have, but also to the history that might bind us, to the things we have said or not said to each other in the past.
G: So, to summarise with a dry answer: can the “society of the mind” and the “levels of understanding” be considered as compasses today?
A: No, I would say not. I would break away from that way of thinking. Perhaps I would keep, in Marr’s case, that slightly more sophisticated structure concerning learning. Let’s think of an example: human beings, particularly children, learn to recognise a cat even from a drawing; all they have to do is look at it once, maybe twice, then they generally manage on their own. I believe that the possibility of disposing of the millions of meticulously drawn examples in the case of supervised learning (of which I have nothing against because they certainly constitute the fuel that makes machines work) could perhaps lead towards computational models that could be considered more valid for achieving the same results.
G: Let’s change the subject. A more general question. Which of these disciplines do you think has an important link with artificial intelligence: philosophy, psychology, neuroscience.
A: As they stand at the moment, I would say neuroscience. The most significant development from this point of view is certainly computational neuroscience. When I was young, there were important figures who were fashionable, namely computational psychologists (Marr is a very good example). They are less popular today than they were at the time. Philosophy, on the other hand, seems to me to have taken a completely different path, I would not stand here and judge whether it is good or bad. Certainly, in the roaring years, those when artificial intelligence was there but didn’t work, philosophy was intimately connected to artificial intelligence. It was not uncommon to meet people (there still are, but with the same colour of beard as mine, [laughs]) who saw a union of the two disciplines.
G: And if you had to choose one particular research topic today, one direction, one topic that you are most excited about and see as promising in the coming years. The famous ‘bone’ to bite…
A: I haven’t been involved in a lot of things in my life, but after a while I like to change bones, let’s put it that way. Certainly the bone I would like to bite the most right now is the one of reinforcement learning. What I’d like to try to do here is something that’s not too different as a way of thinking from what we’ve done in past years on supervised learning. That is, to try to understand in what way it is possible to recount the problems and methods of reinforcement learning within a framework that is a little more respectful of the mathematical subtleties that can be encountered. Again, I feel for obvious reasons ‘demodé’ [smiles], but I would save myself by saying that this is what I like. So I don’t want to say that if you don’t do this you will never fully understand reinforcement learning, but it is something I would like to be able to do. Reinforcement learning seems to me to have a higher structural complexity than the learning I’ve been involved with for several years, which is learning from examples. I have the impression that it goes in a direction that allows us to explain many things, to be able to account for many human behaviours. It’s an interesting paradigm, but I see it as a bit immature. In some ways it resembles what the world of supervised learning was like before mathematicians got their foot in it….
G: Now it’s time for the deeper questions. Let’s talk about consciousness. Here again, I propose a cue, Giulio Tononi, an Italian neuroscientist, who proposes a theory and a beautiful novel called Phi. For him, every living system has a degree of consciousness that can be measured. And he demonstrates this with his studies. This idea of measuring from properties of a system reminds me a bit of Minsky’s idea, the society of the mind. What do you think about that?
A: I bought Tononi’s book because I had gone to hear one of his lectures. It had intrigued me. I can’t say I read it all. I can’t really say how much of it I have read. I think I am a very dry person… In a book like that Tononi showed a face that was also that of a writer, that is, of a person capable of telling stories, of keeping a novel, something I am totally incapable of. I am a reader from time to time, but then I lapse into the detective story, into much drier things. I have survived until now by learning not to ask myself questions that I am sure I will never be able to answer. I willingly ask myself questions I don’t know the answer to, in fact I would say I tend to prefer questions I don’t know the answer to, but I try to devote myself to those for which I think an answer can be found. Even mentioning the word consciousness makes me a little uncomfortable, because I have the impression that I am entering an unknown world, as if I were being asked ‘what happens after earthly life? I am, after all, an atomist…
G: That’s where I fit in! In the next question, the key word is faith. Many scientists profess to be atheists or agnostics and it tends to happen many times that you put religion and science in separate baskets. Do you, as a scientist, see them as two separate baskets? Or is there something that connects them?
A: Well, I have a fundamentalist background, so I also have the animosity of someone who probably no longer believes. Atheist is a bit of a strong definition for me. My conception of life is that life is not mine. In other words, I have a strong conception of life in the sense of the life that exists for example right now on this planet or even for a long time on this planet. My interest and trust in life stems from the fact that there is someone who will continue it after me. Life for me is not only mine, it is not only that of my children or my wife, although of course I love them dearly, but it is also the life of those I do not know, those who have not yet been born and even those who have already died. So, in this sense, this conception does not make me separate science from religion. That is, if this can be considered religion, then I would be happy to see it as my religion. As far as organised religion is concerned, I cannot separate science from religion. What I often say to my friends when we talk is that you cannot overlook the fact that the bar has been raised a little bit since the Middle Ages, you have to come up with schemes that are a little bit more modern [laughs]. I mean I think the explanation has to be a little bit more credible… I don’t even want to hear about something that lasts 2000 years, I don’t know what that means in scientific terms…
G: Let’s conclude with some softer questions. Genoa is an icon, for many it is not just a city, it is something more. What does it represent for you?
A: I don’t know… I would say home! It is the place where I was born, where I will reasonably pass away. I have also lived in other places and had a great time. If I had been asked the same question when I was twenty, I would have answered that there was a much deeper sense of belonging to the city. Now, I feel it much less. I would have no problem moving to another place. I would perhaps have a problem leaving the water, even though I don’t like swimming. I like the water, even not seeing it, but knowing that the sea is nearby is something I enjoy. Boston (in which I have lived very well) is a peculiar city where you can hardly ever see the sea unless you live in the right places, but perhaps knowing that I could walk there played a good role.
G: Speaking of people to come, speaking of the future, let’s conclude with a little advertisement. What would you say to a student to entice them to be part of MaLGa? To choose Genoa, to choose that point on the map…?
A: The attempt that MaLGa is making is to create an environment that is probably a bit of a bubble within today’s climate in which it is possible to deal with issues that are interesting at the moment, are fermenting, are creating interest a bit everywhere, with a style and a defensible, very pronounced attitude. Probably combined with the fact that we are not talking about a place where aggressiveness or the desire to emerge has too strong a connotation [the University of Genoa, ed.] So it’s an attempt to enhance the things we know how to do and at the same time not lose the low-profile identity that distinguishes us. And then things are going particularly well, we are living in a good moment, in the sense that we have had a lot of luck or quite some value in getting a lot of funding.
G: I confirm that! Well, thank you! That’s it!
[*] Tomaso Poggio: Brains, Minds, and Machines | Lex Fridman Podcast #13 https://www.youtube.com/watch?v=aSyZvBrPAyk