Rethinking Machine Vision: New Brain Mapping to Improve AI?


Brain is one of the most interesting parts in an organism. It has been confounding and astonishing scientists with its depths in terms of function and mechanism in equal measure. Thus, it comes as no surprise that even after all these years of research, there are still territories left uncharted.

To better understand brain’s visual system organisation, scientist undertook a study on marmosets, delving deep into whole-brain connectivity. It is important to note that it is the first time scientists conducted research on primates, animals with humans-like brain structure. This implies that sometime in the future, we might be able to transfer the value of this research to the human domain.

As of today, a team of neuroscientists from Cold Spring Harbor Laboratory and University of Sydney have found evidence that calls for revision of the older theory.

Old vs. New

The traditional view was that different thalamic nuclei are either relay nuclei or association nuclei. For instance, according to this view, visual thalamus has lateral geniculate nucleus and the pulvinar. The former relays, latter is responsible for multi-sensory coordination, However, the evidence finds that in specific regions, both parts have same kind of cells. That means sub-compartments may be sharing the same function instead of following a strict labor division. This also means that the two parts might be collaborating in a way previously unnoticed.


This study has, therefore, opened gates to working on the design of artificial neural networks for improving machine vision. As per CSHL professor, Partha Mitra, who is also a senior author on the study, this has potential of rethinking network algorithms. Therefore, it can be safely said, that science is closer to doing away with dated anatomy of visual system. And, by that logic, it is closer to enhanced AI.

Leave a Reply