Neuroscience meets AI
Thoughts on how research in Neuroscience and Artificial Intelligence can come together to understand principles of computation in all neural systems
The research in neuroscience, cognitive science and artificial intelligence (AI) has been intertwined for many decades. While we are going through waves of ideas across these fields being more or less aligned, researchers across these fields certainly have had large influences on each other’s thinking in the past – and will continue to profit from each other’s works in the future. The most recent outcome of these interactions in the emergence of a new subdiscipline: NeuroAI
NeuroAI promises to finally create a more permanent link between neuroscience and AI. While this field is still in the process of defining itself, I believe that it will be in a unique position to study the principles of computation / information processing underlying the functioning of both biological and artificial networks. By finding links between methods and data from both neuroscience and computer science, the field should try to uncover basic principles underlying intelligence in brains and machines.
On this page I want to create an informal collection of ideas and resources in NeuroAI, to give researchers and the public the opportunity to explore and follow recent debates.
If the topic is of interest to you, I discussed many of my ideas in this presentation and interview with Researcher App. Here I go through both a high level overview and also explain some detailed technical ideas of ongoing projects: https://www.researcher-app.com/paper/14619682
Current topics in NeuroAI
While this new subdiscipline is forming, many senior researchers are contributing to review and opinion pieces to give input on what we shoud strive for. One very recent overview piece highlighting the potential link between fields is "Toward Next-Generation Artificial Intelligence: Catalyzing the NeuroAI Revolution" (https://www.nature.com/articles/s41467-023-37180-x). For highlighting potential benefits on the AI / ML side, I find "Deep learning needs a prefrontal cortex" a good overview (https://baicsworkshop.github.io/pdf/BAICS_10.pdf). For highlighting the impact on neuroscience, "The neuroconnectionist research programme" (https://arxiv.org/abs/2209.03718) is a great place to start. In addition, Patrick Mineault is writing on NeuroAI In his blog: https://xcorr.net
Brain-inspired learning in artificial neural networks
Artificial neural networks (ANNs) have emerged as an essential tool in machine learning, achieving remarkable success across diverse domains, including image and speech generation, game playing, and robotics. However, there exist fundamental differences between ANNs' operating mechanisms and those of the biological brain, particularly concerning learning processes. Together with colleagues I worked on a review which presents a comprehensive overview of current brain-inspired learning representations in artificial neural networks. We investigate the integration of more biologically plausible mechanisms, such as synaptic plasticity, to enhance these networks' capabilities. Moreover, we delve into the potential advantages and challenges accompanying this approach. Ultimately, we pinpoint promising avenues for future research in this rapidly advancing field, which could bring us closer to understanding the essence of intelligence.
Building artificial neural circuits for domain-general cognition: a primer on brain-inspired systems-level architecture
In our recent paper we discuss links between recent findings on domain-general cognition in the brain to efforts in building multimodal and domain-general AI models: https://arxiv.org/abs/2303.13651
I also discussed these ideas on this topic with Researcher App. A recording of the interview is available on: https://www.researcher-app.com/paper/14619682
How to get involed in NeuroAI: Join the open-source research group
Research groups working on NeuroAI with a full mix of ML/AI and Neuroscience are still sparsely distributed across the globe. To make it easier for new researchers to first start with NeuroAI or to find new collaborators, OpenBioML (https://openbioml.org) is expanding their open-source BioML research lab with a NeuroAI track. If you want contribute, join the Discord (https://discord.com/invite/GgDBFP8ZEt). There you will find a NeuroAI channel with ongoing discussions on projects. I have been collaborating with them to identify relevant research directions and we would love you to join the research community if you want to be part of our discussions or even help with work on upcoming projects.
You can find the open-source guidelines of OpenBioML here: https://openbioml.org/approach_and_partners.html