I natively assume that there's a "me" which does the "controlling" part of executive function. This utterly fails[1] to predict how (initially promising) neurological interventions change cognition.
Let's look at working memory capacity. I thought that, by stimulating brain areas in special patterns, I could have an external system take the role of "controller", leaving me with more "control" headroom; and in this way, I could use my "will" to hold more items in memory.
For example, I could run software which re-activated something I was thinking about a few minutes earlier, and spend "my" control on meta-analyzing my earlier thoughts. Thus I'd get free, extreme metacognition.
Nope.
On the model I was using, working memory "items" are the coordination of many individual pieces of brain tissue to sit on one carrier frequency; "control" is better thought of as a result of those areas all co-optimizing for sitting on one frequency.
And because of a fundamental harmonic uncertainty, these coordinations are limited to no more than a few tens of WM entries. This feels introspectively like a lack of control-of-stuff-in-working-memory.
From the perspective of any coordination-clique, the stimulation is still cooperative; the same ground-up process occurs, and the "coordination" resource is still consumed.
So the thing which my mind compresses into limited-ability-to-control-memory still happens. I don't get much, if any, extra cache.
(But this sharper map implies that I am able to expand my RAM, by building an index and writing more efficient tooling than what a human brain could reasonably implement. At timescales below ~500ms, my cognition doesn't obviously seem more powerful; but above 10s, and especially for things which take a night's sleep to learn, engineering matters.)
"Control" is the result of lots of local computations done by brain regions; this implies that we can't natively expand working memory. It also helps understand psychosis symptoms in digital telepaths (next post).
Follow-up to both memory posts.
I natively assume that there's a "me" which does the "controlling" part of executive function. This utterly fails[1] to predict how (initially promising) neurological interventions change cognition.
Let's look at working memory capacity. I thought that, by stimulating brain areas in special patterns, I could have an external system take the role of "controller", leaving me with more "control" headroom; and in this way, I could use my "will" to hold more items in memory.
For example, I could run software which re-activated something I was thinking about a few minutes earlier, and spend "my" control on meta-analyzing my earlier thoughts. Thus I'd get free, extreme metacognition.
Nope.
On the model I was using, working memory "items" are the coordination of many individual pieces of brain tissue to sit on one carrier frequency; "control" is better thought of as a result of those areas all co-optimizing for sitting on one frequency.
And because of a fundamental harmonic uncertainty, these coordinations are limited to no more than a few tens of WM entries. This feels introspectively like a lack of control-of-stuff-in-working-memory.
From the perspective of any coordination-clique, the stimulation is still cooperative; the same ground-up process occurs, and the "coordination" resource is still consumed.
So the thing which my mind compresses into limited-ability-to-control-memory still happens. I don't get much, if any, extra cache.
(But this sharper map implies that I am able to expand my RAM, by building an index and writing more efficient tooling than what a human brain could reasonably implement. At timescales below ~500ms, my cognition doesn't obviously seem more powerful; but above 10s, and especially for things which take a night's sleep to learn, engineering matters.)
"Control" is the result of lots of local computations done by brain regions; this implies that we can't natively expand working memory. It also helps understand psychosis symptoms in digital telepaths (next post).
In category theory, this sort of information loss via abstraction is called a generative effect.