Seeking to clarify the basis of AI agent design
A question about Yann LeCun’s path toward autonomous machine intelligence
Why link anatomical structure with cognitive functions as the basis of architecting autonomous intelligent agents?
Figure 2: A system architecture for autonomous intelligence.1
Section 8.2.1: Could this Architecture be the Basis of a Model of Animal Intelligence? Many of the modules in the proposed architecture have counterparts in the mammalian brain that perform similar functions.2
In addition to anatomical structure, what about focusing on functional interconnections by extending the learnings from the Human Connectome Project3 (HCP) and implementing connectomics4?
For example, create differentiable modules that reflect a 7-network or 17-network cortical parcellation.5
7 networks of the cerebral cortex:
Visual
Somatomotor
Dorsal attention
Ventral attention
Limbic
Frontoparietal
Default
Or, does the focus on the brain’s physical structure go well with Joint-Embedding Predictive Architecture (JEPA) and performing predictions in representation space?
Understanding the basis of design might help researchers integrate discoveries and inventions.6
A Path Towards Autonomous Machine Intelligence - Version 0.9.2, 2022-06-27: https://openreview.net/pdf?id=BZ5a1r-kVsf
A Path Towards Autonomous Machine Intelligence - Version 0.9.2, 2022-06-27: https://openreview.net/pdf?id=BZ5a1r-kVsf
The Human Connectome Project (HCP): http://www.humanconnectomeproject.org/
Wikipedia contributors, "Connectomics," Wikipedia, The Free Encyclopedia, https://en.wikipedia.org/w/index.php?title=Connectomics&oldid=1088550419 (accessed July 24, 2022).
Yeo BT, Krienen FM, Sepulcre J, et al. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. Journal of Neurophysiology. 2011 Sep;106(3):1125-1165. DOI: 10.1152/jn.00338.2011. PMID: 21653723; PMCID: PMC3174820. https://europepmc.org/article/PMC/3174820
A Path Towards Autonomous Machine Intelligence - Version 0.9.2, 2022-06-27: https://openreview.net/pdf?id=BZ5a1r-kVsf
p. 1: Many ideas described in this paper (almost all of them) have been formulated by many authors in various contexts in various forms. The present piece does not claim priority for any of them but presents a proposal for how to assemble them into a consistent whole.