Artificial Intelligence

Introduction

Artificial Intelligence, the dream of logical thinking people because it can solve so many problems and would be able to fill a lot of jobs so there is more free time usable for the population. However real artifical intelligence is very hard to create because the more you expect of it, the more complicated it gets. At the moment I am in a state that says that such an AI can solve your problem, what ever it is; just the more you want, the more you've to teach it and the longer it needs to compute a nice answer. Teaching doesn't mean talking to it however in most cases, it's more editing its source code. Also a self learning AI that wants to understand the world, needs a view into the world: sensors - they make the data flow really big and the thing complicated. Because of that the actual goal is more like a search engine of information and logical combination of this / understanding different languages. When you try to understand different languages however, you've to think out side of your own language and the languages you know, so the thing becomes really problematic since you'd have to know all types of languages spoken on earth and still couldn't be sure whether that's your answer.

The Beginning

What does language consists of? You've to know that since you want to comunicate to your AI in some way and most times the most effective one seems to be a spoken/written language. ...
The main inspiration for AKISA was to build a robot that can solve your homework and research for you in things you're interested in.

Won Knowledge and Experiences

- AI systems must be between abstract and exact. They find their way in their possibilities. We ourselfes luckily have the possibiliy of building machines which increase of measurings of nature.

- A perfect AI must be able to learn new principals of learning and gaining information.

AKISA

"Antonios k√ľnstliches Intelligenzsystem - alpha" meaning Antonio(me)s artificial intelligence system, version alpha; is the first project of playing with "intelligent" systems and the try of a system gaining information by itself and working with them so they can answer you questions in their own specific language.

All AKISx projects have the save goal: to reach an AI looking system. Anyways this is the article about AKISA so I'll tell you more about her:

AKISA begun with interpretation of single sentences like you find them in Wikipedia. It was built for the english language because it has no grammatical cases and Wikipedias data set is there the best. But when you really try to understand such a world you'll pretty fast see that there is no simple sentence like SPO used. Very often it's much more complex and sentence structures are combined with each other. Just see my sentences and you'll know what I mean.

Beside the problem of he, she, and it, which are all no direct pointers, the basic interpretation structure is based on static words which will never change their meaning, and flexible words which are defined by sentence contents. The basic principle every knowledge fragment is based on is that Object A -Way C-> Object B when Condition D. This allows a lot of things, still not all and D can become very complicated so I tried to split it into a chance(%) and a regularity(time). Additionally from basic sentences the fact that they exist next to each other, that they belong together is added to the database, the flexible brain of AKISA.

With that principle a simple FAQ is possible and combined knowledge is possible. When there is an interpretable information source, AKISA can learn pretty fast and well new information structures. Still AKISA has no algorithmical or logical core which is needed for intelligence.

Questions are asked by keywords.

AKISA as the start was the biggest part of the project, because it sounds much nicer than AKISB and the ideas could be combined to make it work better.

AKISB

Another name, working with won knowledge, a new project part: AKIS - beta.

The idea of AKISB wasn't easily combinable with AKISA so I decided to make a new subproject of AKIS. So what's new? Better: what is AKISB?

AKISB is based on the principle of actio and reactio, of a brain and a body. For that a synthetical body is created to give big informational inputs, and multiple outputs. The brain has to decide whether doing a thing is good or bad, and whether information can be gained. It's like reflecting over yourself: a pretty complex problem.

The structure isn't that complex through: the system is tick based, so every tick all body parts are asked for new information, the standard importance value is increased, to less important things of the past go gone, possible actions on the input are researched for consequences, and those are evaluated for taking a decision which then is given to the body(output). It sounds easy, but taking decisions and measuring how good sth is, is hard for a computer to understand, especially when it has no experiences, to there is no data for evaluation. So before the system can work, it needs some base data or an explanation of good and bad.

The informational system was based on a short time and long time knowledge base, and inside that words, and objects consisting of words.

AKISC

Back to work with words instead of a synthetical body: AKISC with the idea of implementing feelings(good-bad, importance for itself), and deep mining into data, finding your way from deserts over cities to water, civilisations, humans, arts and technology. All together just a mining system which tries to evaluate a bit.

Complex Neurosystems

A nice topic, a new project, which tries to show the possibilities of self developing simple machines / organisms, for food search, against predators, not walking against walls, ...

And all this based on logical constructions like in our own brains - a funny world where everything develops its most efficent patterns, a population dying and dying in sertain areas just because somewhere else it worked.

This is an image from my testing "laboratory":

hexagons eating moving rectangles

In this scenario, the hexagons can move through walls, the rectangles ("cars") not. The cars have the task to move as much, the hexagons to eat as many cars as possible. When they leave the map too far or are too inefficient, they die.

AKISD

Still in work... trying to understand languages from the unique way all languages have to be formed to allow informational exchange. AKISD tries to understand languages by patterns, and word frequencies. A hard try until you find 340 and 440 ;). Then still it doesn't become easier and you've to research in lots of other languages you never spoke yourself or heard somebody speak before, after questions like: are there always symbols for questions? What's the chinese symbol for a question? What are pointers in this language? Is the order most often like this? Which languages have the pattern XYZ?

Yeah... a pretty complex project when you try to find similarities in all speakable languages of the world, trying to understand its structure and functional words the same as grammatical rules or the best way for gaining information and mining inside the database.