RFR: Left/Right brain in an RL agent

Proposed by Cerenaut and the Whole Brain Architecture Initiative
(What is a Request for Research/RFR?)

The brains of all bilaterally symmetric animals on Earth are divided into left and right hemispheres.  The anatomy and functionality of the hemispheres have a large degree of overlap, but they specialize to possess different attributes.  This principle is poorly understood and has not been exploited in AI/ML.  The right hemisphere is more dominant for novelty, and the left for routine [1, 2, 3].  Activity slowly moves to the left hemisphere as a task is perfected. In this project, we apply that principle to continual RL, where new tasks are introduced over time.  We will create a ‘generalist’ right network that can perform novel tasks while a left network has time to become proficient, providing a more maintained level of competence across new tasks – a critical characteristic for practical agents to operate in realistic environments.

Required Knowledge:
Machine Learning, Deep Learning or some knowledge and willingness to learn.  Must have Python and some experience with PyTorch or Tensorflow.

References
[1] E. Goldberg, K. Podell, and M. Lovell, “Lateralization of frontal lobe functions and cognitive novelty,” Journal of Neuropsychiatry and Clinical Neurosciences, vol. 6, no. 4, pp. 371–378, 1994, doi: 10.1176/JNP.6.4.371.
[2] E. Goldberg and L. D. Costa, “Hemisphere differences in the acquisition and use of descriptive systems,” Brain and Language, vol. 14, no. 1, pp. 144–173, 1981, doi: 10.1016/0093-934X(81)90072-9.
[3] E. Goldberg et al., “Hemispheric asymmetries of cortical volume in the human brain,” Cortex, vol. 49, pp. 200–210, 2013, doi: 10.1016/j.cortex.2011.11.002.

Status: Open

Contact: rfr [at] wba-initiative.org