New ML/Neuroscience Research Projects – ‘Self Organising Architectures’ and ‘Bicameral Brains’

New ML/Neuroscience Research Projects – ‘Self Organising Architectures’ and ‘Bicameral Brains’

We recently published 2 new ML/neuroscience research projects as part of the Request for Research (RFRs)  projects, with WBAI. They’re fascinating topics that have arisen through the relationship with our advisor Elkhonon Goldberg from the Luria Neuroscience Institute.

The projects tackle features of neocortex that are very important for cognition, but not addressed in mainstream ML/AI. The projects will result in computational models that will advance ML as well as neuroscience.

The first is on Self Organising Architectures. For state of the art machine learning (ML), the architecture (receptive fields, their connectivity and integration) is almost always fixed and specified beforehand by the algorithm designer. Mimicking a self organising model should lead to more effective use of resourc

es, better modelling of input and a quantum leap in flexibility for a range of environments and tasks.

The second is on Left and Right Neural Networks – Inspired by our Bicameral Brains. Not only do we humans have brains that consist of two different and specialised hemispheres, but it is a universal feature of symmetric animals. It is clearly an important feature of intelligence. This project aims for improvements in ML by creating a ‘two hemisphere’ algorithm!

As with all of our RFRs, get in touch if you’re interested in learning more or in having a go with one of them. It’s a great opportunity to get involved and make a contribution. Together with Luria and the Whole Brain Architecture Initiative (WBAI) we will be actively advising and guiding the projects.


Also published on Medium.


Comment ( 1 )

  1. ReplyPhille
    About the bicameral one, there is an interesting recent article by Robin Hanson, where he proposes that one half focuses on top-down optimisation and the other half on bottom-up optimisation: http://www.overcomingbias.com/2018/08/separate-top-down-bottom-up.html