We are about as far from it as we were when we lived in caves and banged rocks together.

I'm not so sure about that anymore

Tell your A"I" to generate you an SMP-capable scheduler, aware that each core cluster can be of a different perf and power consumption class. Not a vague explanation of one. Real code, please. Let me know how that goes for you. That is one of 1e9 problems that need to be solved to make a modern OS, not even one of the harder ones.

You don't have to replicate all the insanity that goes on in a modern OS.

Lock threads to a particular core on creation, every core does a Round-robin over all threads. You're 50% there.

Have the core detect congestion, and move yeet threads to another random core when it's congested. You're 75% there.

Use three lists for priorities instead of one. You're 90% there.

Make up for the other 10% with sane design in other subsystems.

We're a long, long way from any model one-shotting an entire OS, but the gains of AI will definitely make it more likely that some team decides to develop a new open source OS where it would have otherwise been impossibly resource-intensive.