The main one is neural plasticity. the damage can be “routed around” and the loss of a column doesn’t create a “hole” the loss of input from the outside doesn’t stop the column from participating.
The other is something animals can’t do and that is plug and play column replacement with or without replacement sensors. new LMs could be swapped in, and the rest of the columns would “bring it up to speed”. That implies the ability to expand the system by adding new LMs with or without sensors to the system, and not need to train them in any way. In essence the vision can be improved by adding LMS and sensors to take in and process more at once, and/or improved sensor preprocessing.
way out on a limb it means an MI could be built a block at a time, learning all the way. Layers could be added for greater abstraction and complexity.