Critical dynamics arise during structured information presentation within embodied in vitro neuronal networks

Area of Science:

• Neuroscience
• Computational Neuroscience
• Artificial Intelligence

Background:

• Neural criticality, near-critical states in brain function, is linked to cognition and consciousness.
• Previous research faced limitations in human and animal models to study neural criticality.
• Understanding the emergence and role of neural criticality remains a key challenge.

Purpose of the Study:

• To investigate the emergence of critical dynamics in neural networks.
• To explore the relationship between critical dynamics and task performance.
• To determine if criticality is sufficient for learning in neural networks.

Main Methods:

• Utilized an in vitro neural network of cortical neurons.
• Trained the network to play a simplified game of 'Pong' to demonstrate Synthetic Biological Intelligence (SBI).
• Analyzed the emergence of critical dynamics in response to structured sensory input.

Main Results:

• Critical dynamics emerged when neural networks received task-related structured sensory input.
• The system reorganized into a near-critical state upon receiving structured input.
• Improved task performance correlated with proximity to critical dynamics.
• Criticality alone was insufficient for learning without consequence information.

Conclusions:

• Neural criticality emerges as a fundamental feature of structured information processing.
• Structured sensory input drives neural networks towards critical states.
• Learning in neural networks requires more than just critical dynamics; it needs feedback on actions.