Biomimicry is certainly is what is being done with the HTM algorithm.
But when it comes to hardware, efforts have been made to use programmable logic gate arrays which usually have a standard architecture, despite their limited plasticity. Shouldn’t HTM hardware be manufactured according to the structure and organisation of the neocortex? A single cortical column does a lot of valuable processing and maybe we can replicate it’s architecture on a practically micro scale.
Does anyone have ideas about designing basic silicon architectures mimicking the function and/or structure of the cortical units?
The approach here should be ground-up.
Are there any such hardware implementation projects or ideas?
It is probably a bit early for this question but nevertheless, it has to be done at some point.
Maybe we can use bacteria to carry out algorithms :v