It’s just that one reason for liking neuron counts is that we have relatively easy ways of measuring neurons in a brain (or at least relatively easy ways of coming up with good estimate). However, as noted, there are a lot of other things that are relevant if what we really care about is information-processing capacity, so neuron count isn’t an accurate measure of information processing capacity.
But if we focus on information-processing capacity itself, we no longer have a good way of easily measuring it (because of all the other factors involved).
This framing comes from Bob Fischer’s comment on an earlier draft, btw.
Thanks, that makes sense. For some reason I read it as a kind of generalisable statement about epistemics, rather than in relation to the neuron count issues discussed in the article.
“the more measurable a metric we choose, the less accurate it is, and the more we prioritize accuracy, the less we are currently able to measure”
Can you expand on this? Is it a reference to Goodheart’s Law?
No, it’s not a reference Goodheart’s Law.
It’s just that one reason for liking neuron counts is that we have relatively easy ways of measuring neurons in a brain (or at least relatively easy ways of coming up with good estimate). However, as noted, there are a lot of other things that are relevant if what we really care about is information-processing capacity, so neuron count isn’t an accurate measure of information processing capacity.
But if we focus on information-processing capacity itself, we no longer have a good way of easily measuring it (because of all the other factors involved).
This framing comes from Bob Fischer’s comment on an earlier draft, btw.
Thanks, that makes sense. For some reason I read it as a kind of generalisable statement about epistemics, rather than in relation to the neuron count issues discussed in the article.