Neuroscience, AI and Dream Machine
a small glimpse into Karim Jerbi’s World
By Jihane Mossalim
When I first heard his talk on creativity, AI and neuroscience during Convergence’s first Scientific Cafe, I was blown away by Karim Jerbi’s mind and way of thinking. I had to interview him. Fortunately for me, he agreed.
Fortunately for you, you can watch his talk during the YlnMn Blue Colloquium, from the Parallel Worlds series. Here is a sneak peak:
"Parallel Worlds" is a free online public experience designed by the Convergence Initiative in collaboration with the Montreal Museum of Fine Arts (MMFA). The experience featured three monthly events that included a virtual tour of artworks, a live online colloquium with a scientist and an art guide, and an online art-science workshop linked to the colloquium. During the colloquium, a scientist and an art guide engaged in a dialogue about visual perception and fine art, inspired by a special pigment representing a color and a subject. During the YInMn Blue colloquium, we had the pleasure of hearing and watching the exchange between the two spokespersons Karim Jerbi and Sylvie Douyon.
Before I jump into what we talked about and all the interesting things he said, I’ll briefly give a preview of what Karim does for a living.
Karim Jerbi does many things. What he is probably most known and recognized for is his work as a neuroscientist. He is the head of the Computational and Cognitive Neuroscience Lab (CoCo Lab) at University of Montreal where his research interests involve the understanding of cerebral connectivity, how the brain works as a network, how the different structures of the brain interact with each other and what happens when the communication between the different parts of the brain stops working. The lab uses a combination of different technologies in order to conduct the research such as EEG (electroencephalogram) and a state of the art MEG (magnetoencephalography) machine to measure the cerebral activity. AI tools are also used, including machine learning to do data mining and classification.
He is also Canada’s research Chair in computational neuroscience and cognitive neuroimaging, the director of the UNIQUE Neuro-AI research center, an associate member of Mila (Quebec AI Institute) and an associate professor in the psychology department of University of Montreal, just to name of few. Yes, just to name a few.
I have no idea how Karim finds the time to fulfill all these positions and projects but it seems to be a recurring trait in many of the neuroscientists I meet; the ability to work on different projects and researches as though their brains are in constant need of challenges and new idea explorations. Have we ever studied neuroscientists’ brains?
Growing up in Tunisia, Karim traveled the world and had the opportunity to study in many countries in Europe and the United-States to finally settle in Montreal, Quebec, a hub for both neurosciences and AI.
He is interested in the cognitive aspects of the brain, the conscience, why and how we do the things we do. For him, the brain is definitely the most mysterious organ in our body; it has an enigmatic side to it which sets it apart from the other, more mechanical organs of the human body.
When asked where his fascination with the brain came to be, Karim looks up and away, remembering his childhood. He tells me that when he grew up, his dad always had books and magazines about the brain, parapsychology, telepathy, hypnosis and the likes. Even though his dad didn’t have a career in academia, the brain was definitely a subject he found fascinating and that fascination clearly got passed on to his son: “I was interested in trying to understand how the brain works since I was very young. It’s not something I discovered later on in University. I must’ve been 10 or 12 when I realized that I wanted to understand more about the subject”.
Will we ever be able to understand how our brain works with our own brain? This is something Karim thinks about, remembering a quote he read in a book, a long time ago: “It said something like if our brain was simple enough so that we could understand it, well with that simple brain, we wouldn’t be able to understand it”. He smiles at this strange and complex paradox of being a neuroscientist, trying to understand the brain with his own brain. He says that perhaps, we will never be able to fully understand how our brain works using our own brain. It remains very mysterious and that’s part of his fascination for this intricate human organ.
This opened up the conversation about his work with Artificial Intelligence. Is AI being used to create more humanlike machines or to get a better understanding of how our brain works? These two directions exist and according to Karim, there are some researchers within the AI community who recognize that we still have a lot to learn about the biological brain and that there are still many things that our brain can do that AIs cannot do; there are things that a baby or a young child can do that the most sophisticated algorithm of artificial intelligence cannot yet reproduce. One of these things include generalization: “to learn a problem and to be able to use our newly acquired knowledge and apply it to a different problem. For example, a child who would learn how to use a hammer in a specific context, the next day, in a new situation, he could take that hammer and use it for something else. For the most part, today’s artificial intelligence algorithms are very, very good at solving a specific problem but as soon as the data or the context changes, even a little bit, there’s a great risk that nothing will work. And that is just one example of what the biological brain can do and where the AI still struggles to keep up.”
With his background in biomedical engineering and his training as a neuroscientist, it is fascinating to see him apply his knowledge in different fields of research, especially when it comes to AI and neuroscience.
Karim is also the proud director of UNIQUE (unifying neuroscience and artificial intelligence in Quebec) which was created in 2019. Its mission: bringing Quebec’s AI researchers and neuroscience researchers together in order to facilitate collaborations and exchange of ideas in both fields of studies.
When asked what the perfect mix of AI and neuroscience would look like, he gives me 3 answers; the first, my personal favorite, is a childhood dream of his: to be able to play one’s dreams in real time, like a movie. The ultimate dream machine. He tells me that research on dreams is being conducted at his lab where researchers are comparing differences in the cerebral activity of people who are able to remember their dreams on a regular basis and people who remember their dreams maybe only once a month. Dreams are definitely on Karim’s radar and I can’t wait to see where his research goes on the subject.
As for his 2nd answer, he would love to see a more ethical and responsible use of AI without biased data. If you train an algorithm with biased data, you will collect biased output. For example, when algorithms were trained to identify faces, it was working very well with white male faces but not so much with women and ethnic faces. Why? Because the algorithm was trained with data which didn’t contain many images of female or visible minority faces. Things are changing but there is still a long way to go. That is something very important to Karim; to work against any form of discrimination and promote equity, diversity and inclusivity everywhere he can, including his own research and workplace.
His 3rd answer is about NeuroAI, the fusion of both fields. The main question is, what is intelligence? Regardless of it being artificial or biological, how would we define it? Which links everything to the notion of conscience and the big question: Can a machine become conscious? At what point do we consider it as being conscious and if so, does it have rights? Would it become better than a human if it had a conscience? Would it be more efficient, more useful? Is it something we want to create? “There are a lot of openings for big questions and opportunities happening with these sorts of interactions”.
It’s obvious that Karim enjoys bridging the gap between his different interests and fields of studies. Art is no exception. Diametrically opposed to very methodical, mathematical scientific measures, art allows for a certain liberty, a certain freedom which Karim truly enjoys. Lover of music and musical instruments, he is sometimes able to bring his artistic interests into the work he does as a scientist. He once worked on a project where he used brain signals and transformed them using sonification, creating interesting ambient sounds. He also worked with some of his students in the creation of the Coco Brain Channel (based on Karim’s CocoLab) with another sonification project which was presented at the McCord Museum during the Printemps Numérique festival.
Karim is a member of the BRAMS laboratory in Montreal, an international laboratory for brain, music and sound research. He also enjoys composing electronic music independently of his scientific research.
Towards the end of the interview I asked what would be the best advice he could give to someone, a young student who would want to follow in his footsteps. His answer: “even if this might sound very cheesy, follow your dreams, really. I mean, we’re all going to die in the end. You could spend your entire life saying that you’ll do something that makes sense money wise but you’ll still die, whereas someone who followed his dream might have had the chance to achieve maybe 30% of his dream which is already great. I would say, if it’s something close to your heart, a childhood dream, try and pursue it. And then when you’re an adult or when you get to your goal, don’t forget that dream you had as a child; people tend to forget and get annoyed at trivial and insignificant daily inconveniences like a canceled reunion. They need to stop and think about that dream they had when they were 15 years old, dreaming to be where they are today.’’
I could have written a lot more about the different subjects we talked about during the interview and I could have interviewed him for days on end; Karim has a talent when explaining concepts and ideas that makes you crave for more. Very generous with both his ideas, points of view and his precious time, I cannot wait to see what his future endeavors will consist of. I’m still hopeful and looking forward to the dream machine.
Parallel Worlds includes a series of live virtual colloquia where brain scientists and MMFA's volunteer guides discuss a topic related to vision, colour and art, an online art-science workshop linked to each of the colloquium themes and a permanent virtual tour that you can visit here!
You can also watch the full colloquium here, in our YouTube channel, with subtitles in English and French.
Parallel Worlds is a free public online experience designed by the Convergence Initiative in collaboration with the Montreal Museum of Fine Arts (MMFA). The experience encompasses a series of events that use the museum's collection to share neuroscience and fine arts knowledge about visual perception with the public. From April to August of 2021, the experience features three monthly events that include a permanent virtual tour of artworks selected for their scientific, historical, technical and aesthetic value. A live online colloquium where a neuroscientist and a volunteer art guide discuss a topic related to one aspect of vision, colour and art. And an online art-science workshop linked to the colloquium. Each online session will be live broadcast on Zoom and other social media platforms. We hope to bring a diverse audience together to explore the advances in visual neuroscience and the impact and influence of art on those advances. Parallel Worlds is supported by the Brain Repair and Integrative Neuroscience Program (BRaIN) of the Research Institute of the McGill University Health Centre (RI-MUHC), Concordia University Faculty of Fine Arts, The Canadian Association for Neuroscience (CAN), and a Knowledge Mobilization grant from McGill University's Healthy Brain, Healthy Lives Program (HBHL).