- I presented soft proof of why convolutions can be used as a foundation for cortical mini-columns.
- I also proposed a novel convolutional architecture according to the thousand brains theory of intelligence.
Does the design for the cortical convolutions improve upon convolutional kernels according to neuroscience and the 1000 brain theory?
Do you think that the existence of this proof and the success of CNNs can serve as weak empirical evidence for the thousand brains theory’s solution to the Binding Problem?
It reminds me of convergent evolution where the evolution of the fields of both deep learning and neuroscience independently lead to similar designs.
Link to article -