summary: A new computational neuroscience study sheds light on how the brain’s cognitive abilities develop and could help shape new AI research.
Source: University of Montreal
A new study offers a new neurocomputational model of the human brain that may shed light on how the brain develops complex cognitive abilities and advance neural artificial intelligence research.
Published on September 19, the study was carried out by an international group of researchers from the Institut Pasteur and Sorbonne University in Paris, the CHU Saint-Justine, the Mila – Quebec Artificial Intelligence Institute and the Université de Montréal.
model who made the cover of the magazine Proceedings of the National Academy of Sciences of the United States of America (PNAS), describes neural development at three hierarchical levels of information processing:
- The first sensorimotor level explores how internal brain activity learns patterns from perception and associates them with action;
- The cognitive level examines how the brain contextually links those patterns;
- Finally, the conscious level considers how the brain detaches from the outside world and manipulates learned patterns (via memory) that are no longer accessible to perception.
The team’s research gives clues into the core mechanisms underlying cognition due to the model’s focus on the interplay between two fundamental types of learning: Hebian learning, which is associated with statistical regularity (i.e., repetition) – or as neuropsychologist Donald Hebb put it. is, “neurons that fire together, wire together” – and reinforcement learning is linked to reward and the dopamine neurotransmitter.
The model addresses three tasks of increasing complexity in those levels, ranging from visual recognition to cognitive manipulation of conscious perceptions. Each time, the team introduced a new core mechanism to enable it to progress.
The results highlight two fundamental mechanisms for the multilevel development of cognitive abilities in biological neural networks:
- Synaptic epigenesis, with Hebian learning at the local level and reinforcement learning at the global level;
- and through self-organized motility, spontaneous activity and balanced excitatory/inhibitory ratio of neurons.
“Our model demonstrates how neuro-AI convergence uncovers biological mechanisms and cognitive architecture that could fuel the development of the next generation of artificial intelligence and even eventually lead to artificial consciousness, Team member Guillaume Dumas, assistant professor of computational psychiatry. UdeM, and a principal investigator at the CHU Saint-Justine Research Center.
He added that reaching this milestone may require integrating the social dimension of cognition. Researchers are now looking at integrating biological and social dimensions into human cognition. The team has already pioneered the first simulation of two whole brains in conversation.
The team believes that future computational models in biological and social realities will not only continue to shed light on key mechanisms, but will also help provide artificial intelligence with a unique bridge to the only known system with enhanced social consciousness: human mind.
About this computational neuroscience research news
Author: Julie Gazelle
Source: University of Montreal
contact: Julie Gaziel – University of Montreal
image: Image is in public domain
Basic Research: open access.
,Multilevel development of cognitive abilities in artificial neural networks“By Guillaume Dumas et al. PNAS
Multilevel development of cognitive abilities in artificial neural networks
Several neuronal mechanisms have been proposed for the formation of cognitive abilities through postnatal interactions with physical and socio-cultural environments.
Here, we introduce a three-level computational model of information processing and acquisition of cognitive abilities. We propose the minimum architectural requirements for building these levels, and how the parameters affect their performance and relationships.
The first sensorimotor level handles local unconscious processing, here during a scene classification task. The second level or cognitive level integrates information globally through long-distance connections to multiple local processors and synthesizes it in a global, but still subliminal way. The third and cognitively highest level handles information globally and consciously. It is based on the Global Neuronal Workspace (GNW) principle and is known as the conscious level.
We use the trace and delay conditioning tasks to challenge the second and third levels, respectively. The results highlight the need for epigenesis through the selection and stabilization of synapses at both the local and global scales to allow the network to address the first two tasks.
On a global scale, dopamine appears to be necessary to properly confer credit assignment despite the temporal delay between perception and reward. At the third level, the presence of interneurons becomes necessary to maintain a self-contained representation within the GNW in the absence of sensory input.
Finally, while balanced innate intrinsic activity facilitates epigenesis at both the local and global scales, the balanced excitatory/inhibitory ratio enhances performance. We discuss the feasibility of the model in both neurodevelopmental and artificial intelligence terms.