By: Liz Do
9 Feb, 2023
From making neurons play Pong, to studying how psychedelics alter states of consciousness, Adeel Razi, a CIFAR Azrieli Global Scholar in the Brain, Mind & Consciousness program, is taking non-traditional approaches to understanding the human mind.
In a Q&A with CIFAR, Razi shares more about his research, and how understanding the human consciousness and brain function could impact future treatment of mental disorders as well as development of smarter AI systems.
CIFAR: You were part of the team who established a proof of principle that neurons in a dish can exhibit basic intelligence by teaching the brain cells to play Pong. Tell me more about the work and your role in it.
Adeel Razi: This is a really cool work that was started here in Melbourne, Australia, in 2019 with a startup called Cortical Labs. This project, DishBrain, is an industry-academia collaboration so my role is to provide academic leadership, computational expertise, including a theoretical framework within which this project operates.
The idea was that we wanted to develop a different sort of computing paradigm, with a biological substrate. We seek to leverage the only known architecture capable of supporting general adaptive intelligence: the biological neuron.
Current AI is inspired by the brain but can we actually harness the prowess of the brain itself? We thought developing a proof of concept — where lab-grown neural cultures, tied to a high-dimensional, multi-electrode array for sensing and decoding neural activity, that can play an arcade game like Pong — would be really neat.
We know that nature has its own way of doing things, and we know that the human brain is so powerful in a way that it can process a lot of information very cheaply. It only requires energy equivalent to powering a lightbulb, and it is extremely fast at processing incoming sensory information. So, the idea was that brain cells can be cheap computational units to develop a computing platform, or biological AI chips. It’s a bit of a wild concept.
CIFAR: In terms of scaling up this work, what do you think it would take to develop a brain-cell powered computing platform?
Adeel Razi: This is truly futuristic sci-fi stuff, the roadmap we have been thinking is 2040. However, in the short term our DishBrain system provides us an excellent opportunity, a testbed living model of (mini)brains. Instead of using inaccurate simulated brain models, we can experiment with DishBrain to test various hypotheses to understand mechanisms of brain functions. For example, one of the things that humans can do rather easily that machines can’t do is learning multiple tasks. This is what we call “lifelong learning,” whereas with AI machines, they suffer from what we call “catastrophic forgetting.” What happens is, you can train an artificial neural network on one task and it can perform it very well, and then you train it for a second task — but then it forgets the first.
So what we’re doing next is a series of experiments to see if these neural cultures can learn multiple tasks without forgetting the first, just like us humans. We will then be able to understand the biological mechanisms that brains use, like replay and neurogenesis, to consolidate and store new memories without losing previous ones. This understanding will in turn help to build better AI systems.
CIFAR: Another interesting face of your work is in using classical psychedelics and computational modelling in order to understand states of consciousness. Tell me more about this work and its potential applications.
Adeel Razi: My lab has three different research directions. The first direction is developing brain-connectivity methods, in conjunction with brain imaging, to understand how different parts of the brain interact with each other. The second is developing neural-inspired AI schemes, a framework of active inference, to understand mechanisms of how the brain performs reasoning, planning and decision-making.
Then the third, as you said, is experimental work using psychedelics. Psychedelics are pyscho–active compounds that can temporarily alter the mind producing altered states of consciousness like ego-dissolution, out-of-body experiences and synesthesia — like hearing colours and smelling sounds. My lab is currently doing Australia’s first, and the world’s largest brain imaging study, where we give Psilocybin (a natural psychedelic found in more than 200 fungi) to 60 people living in a healthy population.
We are interested in the neural basis of psychedelic-induced subjective experiences. When someone is given Psilocybin or LSD, we want to understand what’s actually happening in the brain, in terms of changes in the patterns of neural connectivity, when someone is having, for example, an experience of ego dissolution.
So, we give people Psilocybin, it takes about an hour to get to the peak effect, and then we put them into an MRI machine where we take very high-resolution, real-time, brain images of what’s happening in the brain. And then we use advanced computational methods that we have developed, like dynamic causal modelling, to look at how the information transfer is happening between different parts of people’s brains, and how it’s different from when they don’t have the Psilocybin in their system.
Psychedelics have become very popular recently because of the enormous potential to cure various mental disorders like PTSD, depression and anxiety, to name a few. But, the thing is, we currently don’t know a lot about how these drugs actually work in the brain. So there’s a large gap between understanding what they do in the brain, and the goal of developing better therapies and better ways to treat people.
Our lab is trying to fill this knowledge gap. We also hope that understanding neural mechanisms of altered conscious states would also help us, in the long run, to build better AI systems, connecting with other strands of research on DishBrain in our lab.
CIFAR: Looking ahead, what do you hope to achieve over the next few years in your work? What are the impossible questions you hope to tackle?
Adeel Razi: Our lab is very new so over the next few years, I’m looking to further consolidate the research directions that I have just discussed, by going deeper into developing a better fundamental understanding of various brain functions. Ultimately, we seek to discover if there is a unifying principle, or a set of principles, that would explain the workings of the human mind. Some say it is impossible, but we will see.
Photo courtesy Turner Institute, Monash University
CIFAR is a registered charitable organization supported by the governments of Canada, Alberta and Quebec, as well as foundations, individuals, corporations and Canadian and international partner organizations.