Is this argument robust in the case of optimization, though?
I’d think that optimization can lead to unpredictable variation that happens to correlate and add up in such a way as to have much bigger effects than noise would have.
It seems like it would be natural for the butterfly argument to break down in exactly the sorts of situations involving agency.
Is this argument robust in the case of optimization, though?
I’d think that optimization can lead to unpredictable variation that happens to correlate and add up in such a way as to have much bigger effects than noise would have.
It seems like it would be natural for the butterfly argument to break down in exactly the sorts of situations involving agency.
I expect there’s a Maxwell’s Demon-style argument about this, but I have yet to figure out quite the right way to frame it.