Software > Software

A real enviorment

(1/5) > >>

Sergius:
Just thinking, is there a such thing as a programing envirment that is suited just for modeling the physical enviorment? Most current programing enviorments are suited for analyzing there own internal little world of 0's and 1's.

I believe it would be incrediably intelligent to create an enviorment which is based on input from the external world and its processing. such a framework would redially support all higher math functions calcuales, trige, analitical gemotry ect.

and the variables ect. in the system would essentually be the establishment of the artifical organisms physical shape and propertys, and abilityes.

The rest would be a system of analizing the raw data from sensors

data would be taken in and stored in a special input class which would support the storing and compresssion of large pieces of data, as well as relevance filters.

Sergius:
This came as i thought today of how intelligence works. an intelligence is the product of its enviorment, and thus i have reasoned that reasoning can only truely come from observation and cotemplation of the enviroment.

for instance a developing kitten waits in a room and stores the data in it. i have realised the direction and intrest of the cat is dynamic outliers in the envirment. the cat quickly looses intrest in humans and the moveing plants, as they "always do that" it instead approches the strange and tatically encounters it, puts it in its mouth to see if it is a power sorce. it establishes the relevance of objects.

shouldent this be how an AI should be brought up?

heres the program set up. the machine takes in input from all  its sensors and stackes all of it in arrays of a gereral picture of that second. for example one sec would be a sequence of diffrerent pixel values ( each on a gird) and a clustering of pieces of sound and their directiong on the grid as well, alonge with Sonic input. all on one matrix.

then each pixel would have data in it for the pixel values(IR. visibal, ECT)
225:23,234:23,999
234:2342,221:23,999
and this grid of data would have other things such as sound and its direction attached.  i thought of uniquely placeing the sound on the image as well as part of the enviorment.

then the  60 by 60 or what ever grid of data would be stored data pixel by data pixel in 120 artifical nerons.

the next cycle the old data will be stored in a new set of nerons, which will connect with the others based on the relitive sameness of them, and disconnnect with others that are differennt.

eventually this will be the sub chonsiance of teh machine.
 or something like that

i intend on createing a machine useing a "new" language and this general physolphy. the machine would begin as a drone insided with a CMU cam. in a random section of a normal living room would be a RED bouy with a charger, a positive.

hypothetically the machine would be starteled by people walking and the light sorces, creeping aronund <slowly> (infants dont dart around like nuts.) and establishing what goes on in the room. eventually the data base being built will establish most things in the room as irrevalant ( ill get more to the system for this later) and would eventally see the red bouy as the only intresting thing. thus it would contant it and see it gets energy. the robot by instinct would correlate RED with energy, touching everythng red expecting food. eventually however it will figure the shape of the bouy and know to ignore red shirted humans.


Anyone intrested on working with me on this project?

maverick monk:
sounds beutifull, lol, also, i expect this red human probing baby on youtube

Admin:

--- Quote ---is there a such thing as a programing envirment that is suited just for modeling the physical enviorment?
--- End quote ---

yeap there are a few
search this forum for 'robot simulator' (without quotes)

also . . .
http://www.google.com/search?sourceid=navclient&ie=UTF-8&rls=DGUS,DGUS:2006-11,DGUS:en&q=robot+simulator



--- Quote ---shouldent this be how an AI should be brought up?
--- End quote ---

Learning through experimentation is the only way to learn something new (without being told). Unless the robot already has perfect knowledge of the world, which in reality (outside of labs) it will never have . . .


AI is obviously an unsolved problem, but there is tons of past literature on it that could be useful for you.

My approach to AI is different from the mainstream - Im convinced I can do it with only a few microcontrollers . . . Im currently working on my theories with my newest robot. If any of them work out, Ill write a tutorial or three.

I feel the reason why we havent solved AI is not the hardware, but the software. We got freakin teraflop super computers but havent even figured out insect intelligence yet :P

Something fundamental is being overlooked . . .

Sergius:
Something very fundemential indeed.
ok heres anouther stab at it.


Intelligence Framework

Animalistic
Nomadic and self-sustaning

1. Observe current scene
2. record observations
3. compare past data to current, anything different, unique here?
4. is there anything going on that is unique in the scene,anything different?
5. map locations of outliers and approach them. *
6. observe outlier for response.
7. no response? Look for more new data,

* if something before irrelevant in the situation, such as a wall causes ?disorder?
i.e. vibrations in the body/prevents movement, that is to be maped. Such things are to be avoided and only observed if they prove to be bothersome after avoiding by movement

In an enviorment something isn?t goning to approach or stand next to a wall because a wall is boring. The wall always does the same thing this is how obstacle avoidance works.

 Humanistic:

The human Framework is same to begin with, exept the device is inhibited to observe only certain pre-built in things.

Starting, The human child does the above, but less aggressively. It does not actively seek outliers after observeing a scene, it looks for a match to its first data input, which would be its parent. All data is centered and is connected to the parent and things of and around it.

The human child is dircted by trainers parents ect. to relevant outliers and observational subjects, such as words and letters and toys and tools. These are recorded in memory as bad or good, based on inputs resulting from close proximity to them. The bad and good are still very animinal, anything which disaccoites its matter is bad. Tickeling and such are maped as good as the response from the mother is positive. Generally the feelings it first felt after birth are good, and the tones and data it hears are maped as good and desireable input. Instead of running or avoiding bad, it emits the base cry response- until it knows enoutg to be told not to and understand how and why not to.



2. after the parent exposes it to more and more human things with positives, it more and more connects those human things with posotives The human quickly by base logic establishes the relationship between tool and human.
2a Positives are provided by having intrest in the things first provided by humans to observe.

3a Eventually the device learns to learn and accepts interrupts and is guided to subjects to observe by other humans, as this provided positives.

3b At first in learning the human is taught by the other physically manaplalting or menally tricking the device into doing the task it wants it to.

3cThe human is interested in the positives, the praise.

4 Later the human is able to mimic other humans as it is shown how to mimic by the same above physical movement. The device associates the movement and the human as a ?provider? of data. The taskset and motions are now releveant
5. it is soon told to do things on its ?own? And is left to do tasks by itself or it will receive a negative. The device begins doing unnatural tasks.

6. after this is accomplished, it is shown to teach other devices. It passes on knoladge and data to the next.
Age=7-10

the device is finished with this and is now taught to observe non-human things for input, while abideing to human standards of observation. Because mimicing is now connected with observation in the device, it begins using its learning methods to mimic actions in nature, events, and other devices, and in technology which is other results of this.

8. creativity is born in the device, although is limited to human conformers and learning methods first instated.


Navigation

[0] Message Index

[#] Next page

Go to full version