- Forum
- General C++ Programming
- EvE: A different AI (?)
EvE: A different AI (?)
To begin with, I'm probably way over my head, and have my asbestos suit at the ready as these threads seem to generate flame wars. That being said there's a wealth of knowledge on here and giving it a try seems like the thing to do as the intention is to make it completely open source.
Hardware : 10 parallella boards & to begin a 3TB linked to store the presumably huge database that will be generated.
The Plan : This may or may not have been tried before, but I'm going to go on the assumption that at least this implementation is different. I'm not trying to model a human, cat or any other form of existing brain, the intention is to create something... Different. I feel the answer to true 'intelligence' is to start from scratch. Life is random, it evolves, so start with something simple like an amoeba, a single cell, that appears random and acts randomly based on simple rules, eat, divide etc.
To accomplish this we need something irrational yet ordered, so the mathematical model will be based on the most irrational number I know (pi) applied to the 8th power 'mandelbulb' 3d fractal equation. With numbers pulled from within this there should be the right amount of order and randomness.
This is the core of EvE, applying rulesets beyond that may simulate learning and allow the rewriting of code on an evolutionary level, learning ways of doing things that don't work, and stumbling on ways that do. Of course this will be nowhere near efficient, but that is not how life works.
Rulesets should be formed as plugins, representing different parts of the brain, (language centres, visual centres, cryptography centers) and be able to be called by any other plugin, if required by the random action of the core (I wonder what this sound looks like?)...
Thoughts? Or am I barking up the wrong tree?
Hardware : 10 parallella boards & to begin a 3TB linked to store the presumably huge database that will be generated.
The Plan : This may or may not have been tried before, but I'm going to go on the assumption that at least this implementation is different. I'm not trying to model a human, cat or any other form of existing brain, the intention is to create something... Different. I feel the answer to true 'intelligence' is to start from scratch. Life is random, it evolves, so start with something simple like an amoeba, a single cell, that appears random and acts randomly based on simple rules, eat, divide etc.
To accomplish this we need something irrational yet ordered, so the mathematical model will be based on the most irrational number I know (pi) applied to the 8th power 'mandelbulb' 3d fractal equation. With numbers pulled from within this there should be the right amount of order and randomness.
This is the core of EvE, applying rulesets beyond that may simulate learning and allow the rewriting of code on an evolutionary level, learning ways of doing things that don't work, and stumbling on ways that do. Of course this will be nowhere near efficient, but that is not how life works.
Rulesets should be formed as plugins, representing different parts of the brain, (language centres, visual centres, cryptography centers) and be able to be called by any other plugin, if required by the random action of the core (I wonder what this sound looks like?)...
Thoughts? Or am I barking up the wrong tree?
That's easy, the stubborn persistence my brain has for thinking in English instead of C++.
As for the program there are many hurdles. First and foremost getting the thing to execute in parallel across the cluster, while it may be easier to get it to run on one core, getting it natively parallel and scalable to as many cores as are in the cluster seems to be the appropriate first step.
As for the program there are many hurdles. First and foremost getting the thing to execute in parallel across the cluster, while it may be easier to get it to run on one core, getting it natively parallel and scalable to as many cores as are in the cluster seems to be the appropriate first step.
What research have you done to say this is the 'right amount of order and randomness?'
Why the 8th power mandelbulb as apposed to a 6th power, or 2nd power? Why not a super-Newton or a cos-sin formula?
Way too much. Rulesets should be dealing with one very basic function - Survival.
Why the 8th power mandelbulb as apposed to a 6th power, or 2nd power? Why not a super-Newton or a cos-sin formula?
| Rulesets should be formed as plugins, representing different parts of the brain, (language centres, visual centres, cryptography centers) and be able to be called by any other plugin, if required by the random action of the core (I wonder what this sound looks like?).. |
Way too much. Rulesets should be dealing with one very basic function - Survival.
The mandelbulb formula applied to the 8th power produces a very 'organic' 3d shape, the exterior looks like plant biology and the interior looks like the interior of animal life, optically it's pretty and the fractal nature of the thing produces a very complex pattern mathematically, in 2d it kind of animated in a similar way to a cell. I guess that's just my preference for the thing, and perhaps to give it a form to contemplate should it ever become self aware, something complex enough to keep an AI occupied for a while as the patterning potentially goes on forever. Maybe that's thinking too far ahead but it seems appropriate to me to give an AI as much to mull over about it's own nature as possible... That deals with the fractal part, should it find pi from that it'd have another huge challenge to analyze as that number might just go on forever also... And should it get to the circle from that then it might just be wondering how a simple geometric shape can cause all that complexity. Thinking maybe too far ahead... Probably.
I'd disagree with the notion of basic survival, it'll have to crop up at some point, but hardcoding it in might be a bad idea to start with, as (and you can call me nuts if you choose) it may end up smarter than we are and decide to survive at the cost of the creators.
Perhaps a little too philosophical, but it makes sense to me.
I'd disagree with the notion of basic survival, it'll have to crop up at some point, but hardcoding it in might be a bad idea to start with, as (and you can call me nuts if you choose) it may end up smarter than we are and decide to survive at the cost of the creators.
Perhaps a little too philosophical, but it makes sense to me.
Topic archived. No new replies allowed.