This is confusing genetic markers for intelligence with intelligence. Building half a bridge is not useful, and individual markers don't stand alone.
I do believe that the "imbalance"-hypothesis he presents is a bit vague and proposes several imbalances according to different (and possibly orthogonal) models for intelligence, making it seem a bit unfalsifiable to me. Who knows. Behavioral neuroscience based on MRI(or similar modalities) is a giant sandcastle anyway, best not stack those results too high.
The short version, to me, was that general intelligence is based on several cognitive axes working together. I believe the paper lists 3-4 most of which were separable, like verbal scores, mental rotation tests, etc. And with Autism, you have 1-2 of those axes working above average but the rest not. To falsify, you could break down IQ tests for individuals with autism and compare some areas with the general population and also with each other. If they match the expected pattern, they may be onto something, otherwise the theory is incorrect.