Many people will discard the connectivist approach because it's not interpretable. And the first logical choice for them will be the symbolic approach.

They would argue that even though Deep Learning research can be incrementally improved, but at any stage of its development it's not possible to explain what is going on inside of the algorithm. Of course, we see numbers and matrix multiplications, and other linear and nonlinear transformations, but what is the logic behind it?

The same question can be asked about DNA. Why do we transcribe DNA? We don't understand why it's organized the way it's organized. We don't know what effect any alteration can cause and that's why we ban genome engineering. Many researchers little by little statistically guess what some regions do.

If we cannot understand how a chain of proteins can affect the development of our body, and resistance to deceases, then why don't we stop studying that?

If we want to understand how to treat brain diseases we need to understand how neurons work, and how they can be altered, replaced, or overwritten.

Intervention in the brain will require new machines that can scan brain activity on the neuron level, and measure the strength of synapse connections. The machine that can analyze the brain remotely. That doesn't require cutting neurons out of the skull or other ridiculous procedures.

Even when we have all the connections and strengths that together determine pathways, even then we are far from understanding, everything just random spikes of activity. The same as understanding our genome. We know what region should be promoted (?) but what alterations can we make?