Is a Free/Wild Imagination necessary for AGI

[Anonymous No.9498178]

Intelligence requires a wild imagination that drifts through an immense amount of possibilities. I think this might almost necessitate a personality for AGI. Personality being defined as a variance of architecture. Meaning that it will benefit from changing itself or having a variety of instances with variability similar to in humans.

If we understand imagination properly as the search for viable goal targets which can reduce search times to 2^(N/2) from 2^N without. In the search of interesting goal targets and rotating through an intense amount of imagining to find new novel ones an AGI will need wildness/personality.

You can see this in the relationship between schizophrenia and intelligence. The random thought pattern production in our minds is related to imagination and searching for novel imaginary solutions. Once an imaginary solution is found the algorithmic complexity of arriving at it is again impossibly faster than without the goal target due to "working backwards" and "working forwards" simultaneously with the intention of meeting the two solutions. In fact it's impossible to do anything new without such a goal state in mind.

Schizophrenia (wild imagination) link with Intelligence (problem solving) is not a bug. You have to imagine some imagination capable system with ordinary input, very little will come whereas the same system with a wild imaginative input function will begin creating all sorts of things.

AGI should be explored with the concepts of schizophrenia, dreaming, and imagination in mind as useful tools for "goal state" searches.

[Anonymous No.9498229]

early investigation basis vectors for intelligence, hopefully all generally orthogonal

Raw Brainpower
Schizo Rating
Goal Architecture


Raw Brainpower: Something brute like # of neurons, health of neurons, etc
Schizo Rating: the variability of input to the brain via imagination. Likely bad on both the low end and high end, between nothing interesting and nonsense there is a sweet spot for intelligence.
Goal Architecture: The ability of the intelligence to use AGI algorithm architecture, ex in humans is ability to imagine things to set goals, reason etc.

These three basis vectors should be good, but there must be more investigation into networked/distributed intelligence (emergent) versus singular identity intelligences (conscious)

There are likely more important foundational things like what the intelligence uses as basic fundamental foundation. For instance a visual spatial reasoning intelligence could use simple 3d space and a simple newton physics model. Whereas another could be based entirely in the area of languages, sound, etc. Humans for instance base most of our reasoning off hijacking our visual or language processing centers which affects deeply how we think. These mediums can change though such as possibility of learning math or computer science which changes how people look at problems or imagine solutions.

[Anonymous No.9498715]

>>9498178
>If we understand imagination properly as the search for viable goal targets

That's not evenly nearly the full extent of imagination. Imagination is what recognizes forms or concepts.

[Anonymous No.9498809]

>>9498715
>>9498715
Let's take these concepts

Environmental input -> processing -> visual discernment of objects (ex just cycle through objects and look at them, floor, ceiling, screen, etc)

for imagination we can close our eyes and cycle through any number of objects.

If we then create a new object with simple vector math, ex: Car + wings. we can process this quite fast. Car + tank turret.

Now do the following simple experiment

1. Look around your environmental input and discerning objects. Ex: Ceiling, floor, hand mouse, etc
2. Now mix in the car+wing, car+tank turret in the middle of the cycle randomly.


I think it's pretty obvious the same discernment filter is applied when creating the imagination object as we use to discern if we are seeing a car from environment.

So the key aspect of imagination is duality. Not just input from environment but the ability to send a vector to our visual system and get out an imaginary object we can visualize.


The full extent of an AGI is more complex, but from simple to complex.

[Anonymous No.9498815]

>>9498809
>2. Now mix in the car+wing, car+tank turret in the middle of the cycle randomly.

What are you actually trying to say? Like imagine something and look at real things at the same time? Or imagine it being in front of you? Drop the vector stuff and try to put things into meaningful terms.

>So the key aspect of imagination is duality.

That I agree with.

[Anonymous No.9498884]

>>9498815
Okay, to understand this. Look forwards. In your vision are multiple objects hopefully like a ceiling, wall, desk. Your brain unconsciously will discern these things and have them ready for you. Ex. "look at ceiling". the ceiling you are seeing is immediately available unconsciously.

Just go through the various things in your vision cyclically bringing them to your conscious attention.

ex. Look at Ceiling, Look at floor, look at wall

Now mix in an imagination process. Instead of using your unconscious discernment algos it will use a creationary algorithm.

So for instance the process -> floor, ceiling, car with wings, wall. In this process you will use two distinct "algorithms".

[Anonymous No.9498898]

>>9498884
It's a silly experiment but the point is that your environmental input is all ready for you. The creation of a visual imaginary object takes a long time relatively. The thing is it hijacks the same brain areas that do the visual discernment for visual creation of the imaginary object.