[ExI] 1DIQ: an IQ metaphor to explain superintelligence

[Ben Zaiboc]

On 02/11/2025 19:31, bill w wrote:

 > So - we will never map the brain because its topography, if you will 
allow
 > the metaphor, is constantly changing. bill w >


Depends on what you mean by "map the brain". The changes a brain 
undergoes aren't random. And while we'd have a hard time unravelling all 
the factors that influence it's changes of state, we can predict that 
exactly the same brain model, under exactly the same conditions, given 
exactly the same inputs, would produce exactly the same results. If you 
wanted to recreate the movements of the ball in a pinball machine, would 
you carefully measure each and every change in direction and velocity of 
the ball on multiple runs, or would you measure it's weight and shape, 
then model all the channels and flippers in the machine?

The brain's gross structure hardly changes at all, even between 
individuals. It's at the microscopic level that changes are happening 
all the time. The large tracts of white matter, the global and local 
connections are all pretty fixed once someone reaches adulthood. It 
should be relatively easy to map a generic human brain. Less so to map a 
specific individual one.

The way I sometimes think of it is like a river. While the individual 
molecules of water, and the patterns of flow are dynamic, it all happens 
within the constraints of a fairly static river bed and banks, channels 
and obstacles, etc. We don't try to map the transient features, but 
mapping the long-lived ones works well. If you re-create a river's 
banks, bed, etc., then pour water into the model, you'll get the same 
transient features as the original. The better the static model, the 
more accurate the flow and eddies will be.

Limited analogy, I know, and it ignores things like the sand on the 
river bed being changed by, and changing, the flow of the water, but 
going back to the actual brain, if we can accurately map the connections 
between neurons, the types of neurons, the dendritic spines, synaptic 
positions and weights, as a snapshot in time, then create and activate a 
model of those things, it should behave the same as the biological brain 
(i.e. produce the same mind).


 > scanning a brain at one point in time tells us little about what 
changes it will undergo in the next second, much less next week


That's true, but it doesn't matter. At least not if your aim isn't to 
predict what changes it will undergo in the next second, or week.


 >>> You want to understand our minds? Make the workings of the unconscious
 >>> conscious - and that's just a start. Why did the impulse go to point B
 >>> when it left point A rather than to point C? And then trace all the 
points
 >>> in between entering the unconscious and resulting in some idea or 
action.
 >>> And explain each one.


I don't think that will be necessary. Understanding our minds and 
understanding our brains are different tasks. Something can be 
understood in many different ways. For the purposes of re-creating our 
minds in a different substrate, for example, we don't need to be 
concerned with concepts like the unconscious, we need to re-create a 
structure that will behave in the same way. We don't even have to 
understand why it does that, as long as we capture the features that do 
produce the same behaviour. To go back to the river analogy, you don't 
need to understand fluid dynamics, you just need to know where to place 
the rocks. And maybe what shape they are.

So I think that "understanding", on its own, is not a very useful 
concept. There are many kinds of understanding. We need to specify the 
scope of the understanding, and its purpose, otherwise it's like having 
a goal of 'travelling', without saying where.

The main thing to consider is that a dynamic process (the mind) is 
created by a physical system (the brain) in a particular general 
configuration. Constantly changing inputs will create constantly 
changing internal states, but the structure remains the same (within 
limits. There's a scale at which things become interesting, where the 
dendritic spines and synapses are changing, but that could be viewed as 
equivalent to, say, logic gates opening and closing, capacitors 
charging, etc.)

All this would be created in software anyway, so a model of the brain 
can be as changeable as needed. Moreso. We can't change the way the 
corpus callosum is wired, for example, in a biological brain, but we 
could in a simulation.

I think we will only begin to understand our minds after mind uploading 
is realised. It's not a pre-requisite for uploading. Being able to 
recreate the physical structure of the brain (to a currently-unknown 
level of detail) is, though. This probably requires far less 
'understanding' than you'd think.

-- 
Ben