This is what I’m getting at. I don’t see a logical step between raw computation ability and the specific information processing that humans do that must be capable of reaching these limits. And the fact that you reference these operations in flops might be missing the point — it might be that we have to use a different unit of measure to describe the way organic brains or AI brains actually compute; right now flops and state changes seem a little too apples vs oranges to me.
The other issue I take with arbitrarily lowering the energy requirements for a brain (usually so a simulation argument has any ground to stand on at all) is that a brain is an I/O machine, and it is unclear to me whether the correct inputs and outputs could also be simulated with such high efficiency. Sure there are arguments about compression, rendering, etc… but how do we know an AI brain would react the correct/expected way under these energy saving methods?