Neuro Excellence Series: Autism, Synapses, and AI: The Neurodiversity Lens We Can't Afford to Ignore

Human evolution didn't "finish" with fire, stone tools, or even the printing press. Recent findings on layer 2/3 intratelencephalic (IT) neurons, which are the cortical relays that let the brain talk to itself, have shown that our species has always paid for new cognitive capacities with widened variance. Those shifts expanded language and abstraction, but they also raised the probability of autistic trajectories

That's the price of plasticity. You get more possibilities, more variance, more edge cases at the cost of a neat, uniform outcome. Evolution does not optimize for averages; it optimizes for adaptability

Enter AI. Each day, billions of micro-interactions with large language models, recommendation engines, and synthetic interlocutors act as environmental inputs on our cortical wiring. Some argue this is accelerating human "brain evolution" and that extended cognition with AI augments memory, speeds pattern recognition, and nudges synaptic efficiencies. Let me come back to those later

Here's the parallel I draw: Just as shifts in IT, neurons widened neurodevelopmental variance, AI interactions are widening cognitive variance at the societal scale. Some people are demonstrably upskilling, leveraging AI as scaffolding to climb higher. Others are offloading so completely that skills atrophy

Synapses don't lie. In the same way, autistic gene expression differences alter dendritic integration and timing; our constant prompting and scrolling shape the frequency and density of synaptic firing. Tools that compress attention into micro-bursts restructure the temporal dynamics of learning. The substrate of culture is being tuned, faster in some domains, thinner in others

So are we evolving faster with AI? Yes, but not in a single direction. We are widening the distribution again. A minority will use AI to extend their frontal lobes outward into the infosphere; another minority will hollow out. The rest will oscillate in between, depending on context, access, and discipline

This is why my ethics stance is unwavering. AI must be designed as augmentation, not substitution. If we model our future on deskilling and dopamine loops, we induce cultural atrophy. If we design for co-agency, transparent memory, accountable cognition, and human-aligned scaffolding, then we extend what evolution already began

The cortical story reminds us. Plasticity always carries risk. But risk can be navigated if we name it honestly. The choice is ours. We can embrace AI as an amplifier of variance that requires responsibility. Or. Let it drift into another force of passive deskilling

My conclusion is simple. Neurodiversity is the cost of progress, and cognitive sovereignty is the safeguard of the future. AI does not have to make us dumber. It can, if designed ethically, carry the same spark that once re-tuned our IT neurons and gave us language itself