I'm not clear on what counts as innate. You say colour vision is - but doesn't it require learning by being exposed to hues? Just like kittens who aren't exposed to vertical lines haven't learned to see vertical lines...
Does it count as innate if its an ability that everyone or every animal gains under normal conditions? Because learning by training to see edges in all angles and discriminate hues well is something (almost) all of us will learn.
If so, that begs the question of how to view English if all other languages go extinct. Or if there's were a small group of humans early in history and they all spoke the same language. Completely innate?
Also, LLMs have to learn too, right? They do have some innate biases and learning capabilities (by virtue of their architecture) but they have to be trained.
I agree, it's hard to draw a sharp distinction between what's innate vs caused by environment--at the extreme, you could claim it's all environment, since nothing will develop in a literal vacuum in space (or all nature since nothing will develop without genes). To try to sort out how much of a trait is attributable to one or the other, we have to assume variations in the environment (and genes) that conform to some distribution, but you're right that there's nothing special about any specific distribution we choose. I like your example of English being innate if we changed the distribution of the environment so the presence of English was a constant!
I think we're stuck with some fairly fuzzy definition of, like you said, "normal conditions". I think that maps roughly onto the question we want to know, though--we want to understand why some people are different than others, or how some mechanism develops, or how to fix it if a mechanism doesn't develop The story is richer than just "environment" vs "genes", but looking at the variance attributable to one or the other in what we take to be our normal environment is a reasonable heuristic for getting a very fuzzy picture of the mechanism we're trying to understand.
You might like the latest "Revisionist History" podcast by Malcolm Gladwell on Face Blindness. (Just be sure to have the ability to fast-forward through the many ads.)
Fantastic essay!
I'm not clear on what counts as innate. You say colour vision is - but doesn't it require learning by being exposed to hues? Just like kittens who aren't exposed to vertical lines haven't learned to see vertical lines...
Does it count as innate if its an ability that everyone or every animal gains under normal conditions? Because learning by training to see edges in all angles and discriminate hues well is something (almost) all of us will learn.
If so, that begs the question of how to view English if all other languages go extinct. Or if there's were a small group of humans early in history and they all spoke the same language. Completely innate?
Also, LLMs have to learn too, right? They do have some innate biases and learning capabilities (by virtue of their architecture) but they have to be trained.
Thanks!
I agree, it's hard to draw a sharp distinction between what's innate vs caused by environment--at the extreme, you could claim it's all environment, since nothing will develop in a literal vacuum in space (or all nature since nothing will develop without genes). To try to sort out how much of a trait is attributable to one or the other, we have to assume variations in the environment (and genes) that conform to some distribution, but you're right that there's nothing special about any specific distribution we choose. I like your example of English being innate if we changed the distribution of the environment so the presence of English was a constant!
I think we're stuck with some fairly fuzzy definition of, like you said, "normal conditions". I think that maps roughly onto the question we want to know, though--we want to understand why some people are different than others, or how some mechanism develops, or how to fix it if a mechanism doesn't develop The story is richer than just "environment" vs "genes", but looking at the variance attributable to one or the other in what we take to be our normal environment is a reasonable heuristic for getting a very fuzzy picture of the mechanism we're trying to understand.
I wonder if it's possible to train babies' pattern recognition using something similar to the dalmatian image.
You might like the latest "Revisionist History" podcast by Malcolm Gladwell on Face Blindness. (Just be sure to have the ability to fast-forward through the many ads.)