A compromise between the two would be to identify powerful abstructions over nodes and nets that are equivalent to symbol manipulation circuits. Self assembly a

A compromise between the two would be to identify powerful abstructions over nodes and nets that are equivalent to symbol manipulation circuits. Self assembly and self modification are certainly not strange to symbolic logic, but this doesn't work so well for most programmers. So we need more geniuses and they need to tailor neural nets to help them build symbolic logic (if only as intermediate debugging output).