Artificial Intelligence Project

I have created the thread VB Strong A.I. Project too but i got rather bored with people thinking that a dictionary is AI so i will create this one instead.

I believe you are all familiare with the workings within the CyCorp ?
How many of you think that is AI? I guess many of you do, many more experienced AI developers think so.

What the CyCorp (roughly) is that they are giving their ALF a bunch of words and some grammar knowledge. They make sure it has precisly enough information to "intelligently" answer a particular question. Is this AI i ask you?
The answer is No, it is not AI. If they were to start it with NO data then it would be AI. "But that is impossible" many tell me.
Well, of course it is impossible, otherwise normal mortals could solve it.


I hope all of you have seen a simulation of a bird flock started with a handfull of parameters. (etc velocity, distance to walls, distance to other bird(s) )
If you have not seen one, then do so.
Is that AI? No, it is not AI becuase it has no real reason to be intelligent. By simply following parameters intelligence is not born. The parameters should be a assistance to intelligence. (i am rather known for vague explenations

Please Register or Log in to view the hidden image!

)


The ability to speak is indeed important but i have solved it.
In The Republic (our Artificial world for our artificial lifeforms (or ALF's) (named after Plato's book "The Republic") ) we start our Artificial Lifeforms with simply 5 parameters (ill not say which here), a goal (obtain food, wood, water). The database is completely empty. The Code of course does not tell it anything.
These 5 parameters cover the sensories of Vision and Touch. Audio is of course also integrated but as it develops the human language i am not sure at this point if one more will be neccesary or not.

You might think it is a trivial task to understand how to use an Axe but it is not. (yeah, i forgot. The data on how to use certain items does not exist either).
So, how do you understand that the wooden handle on the Axe really is the wooden handle and that the sharp, metal cutting tool should be swung ? ? ?
I think the answer is, that is impossible If the AI is started without any data. But, i have solved it ladies and gentelmen.
Even if it is rather simple how to use a Axe with my complex system and DB the Bow is still rather difficult since it really makes no sense to fire the arrow. But my system includes that aswell.
I guess i am boring you with this data, most are interested in the ability of speech. Well, i have solved that too. By interacting with the "real" world it will gain some data.
Empricial Parameter is set to 3. So if 3 things suggest something (and of course if zero things can counter it) then it is true. This value is of course dynamic and as data suggests that this value is inaccurate the ALF can change it for the better.
Well, if it has a question such as "What color is a Blue Ball?" and has 2 previous questions that suggests that if "What" and "?" is in the same question then "is" is "=". "a" = "1". "color", "Blue", "Ball" are already sorted out by the DB system.
The grammar suggested above is of course not true in every case but remember that it has a Empicial value of 3 so it will take it about 3000 questions before it has gotten most of our grammar right. Then it can talk, it can relate to data from the virtual (or real) world.

So, is this AI? I sure think so. Instead of giving it the data neccesary and then asking it "friendly" questions (which it of course can answer) we are giving it no data and giving it questions which it cannot answer, just so that it can learn from it.

Yes, i have removed some crucial information so that i won't have any opposition. But i will gladly provide the information to anyone who mails me.

I have created the thread VB Strong A.I. Project too but i got rather bored with people thinking that a dictionary is AI so i will create this one instead.

I believe you are all familiare with the workings within the CyCorp ?
How many of you think that is AI? I guess many of you do, many more experienced AI developers think so.

What the CyCorp (roughly) is that they are giving their ALF a bunch of words and some grammar knowledge. They make sure it has precisly enough information to "intelligently" answer a particular question. Is this AI i ask you?
The answer is No, it is not AI. If they were to start it with NO data then it would be AI. "But that is impossible" many tell me.
Well, of course it is impossible, otherwise normal mortals could solve it.


I hope all of you have seen a simulation of a bird flock started with a handfull of parameters. (etc velocity, distance to walls, distance to other bird(s) )
If you have not seen one, then do so.
Is that AI? No, it is not AI becuase it has no real reason to be intelligent. By simply following parameters intelligence is not born. The parameters should be a assistance to intelligence. (i am rather known for vague explenations

Please Register or Log in to view the hidden image!

)


The ability to speak is indeed important but i have solved it.
In The Republic (our Artificial world for our artificial lifeforms (or ALF's) (named after Plato's book "The Republic") ) we start our Artificial Lifeforms with simply 5 parameters (ill not say which here), a goal (obtain food, wood, water). The database is completely empty. The Code of course does not tell it anything.
These 5 parameters cover the sensories of Vision and Touch. Audio is of course also integrated but as it develops the human language i am not sure at this point if one more will be neccesary or not.

You might think it is a trivial task to understand how to use an Axe but it is not. (yeah, i forgot. The data on how to use certain items does not exist either).
So, how do you understand that the wooden handle on the Axe really is the wooden handle and that the sharp, metal cutting tool should be swung ? ? ?
I think the answer is, that is impossible If the AI is started without any data. But, i have solved it ladies and gentelmen.
Even if it is rather simple how to use a Axe with my complex system and DB the Bow is still rather difficult since it really makes no sense to fire the arrow. But my system includes that aswell.
I guess i am boring you with this data, most are interested in the ability of speech. Well, i have solved that too. By interacting with the "real" world it will gain some data.
Empricial Parameter is set to 3. So if 3 things suggest something (and of course if zero things can counter it) then it is true. This value is of course dynamic and as data suggests that this value is inaccurate the ALF can change it for the better.
Well, if it has a question such as "What color is a Blue Ball?" and has 2 previous questions that suggests that if "What" and "?" is in the same question then "is" is "=". "a" = "1". "color", "Blue", "Ball" are already sorted out by the DB system.
The grammar suggested above is of course not true in every case but remember that it has a Empicial value of 3 so it will take it about 3000 questions before it has gotten most of our grammar right. Then it can talk, it can relate to data from the virtual (or real) world.

So, is this AI? I sure think so. Instead of giving it the data neccesary and then asking it "friendly" questions (which it of course can answer) we are giving it no data and giving it questions which it cannot answer, just so that it can learn from it.

Yes, i have removed some crucial information so that i won't have any opposition. But i will gladly provide the information to anyone who mails me.

We at SC & W Labs are looking for :

• A Neural Net programmer (preferebly 30+)
• A Mathematics and Physics bachlor (or higher) who can help us with Physical Enumerations and matemathics for our Fiscal Value and Exponential Fiscal Value Algorithms
• Anyone who consider himself/herself to be truely intelligent (not school grades), true intelligence (the ability to think outside the box

Mail me Rikard Svensson, CEO of SC & W Labs at svenssonrikard@spray.se

 

Log in or Sign up to hide all adverts.

maybe i should had said. Any detailed data you might be interested in will not be disclosed here.

 

Log in or Sign up to hide all adverts.

Anyone at all ?

 

Log in or Sign up to hide all adverts.

There is no general agreement on the definition of AI. Many researchers are already tired of the endless discussions about what is and what isn’t AI (/who has a "better" definition). The more interesting question is how to design systems which would be able to solve complex real world problems. And that question cannot be well answered briefly.

Just code it. I really believe it will do more for you at this point than long discussions with people who already played with this kind of stuff.

 

Have you heard of Chineese Room Argument,with relation to Turing Experiment?

bye!

 

yes.

But i have also read that many are leaving the field of AI because it was too hard for them. The ones that are left apperently invent things as the "Chinesse Room" because they cannot solve the real problem.

They are not even working on AI. If you think so then i think that the outcome of my AI would be better than a world dominated by man.

 

:m:

Please Register or Log in to view the hidden image!

:m:
Baal Zebul../ Rikard Svensson:

Sorry to put cold water in your blood, but this does not
excatly sound like something new.

It just languale evolution simulation,
of how to evolve language in a population of
objects with no previous experience/data.
This is rather trivial now, as many reseatchers have
duplicated the work. But non the less is it true, and it is beeing done.
And I guess it can be seen as a form of very low REAL AI,
as it is built bottom up from nothing.
The agents simply learn form random interaction with each other and the enviroment. Forming a consensus of a comon conseptual language describing objects and their physical relations.

A good reference is:

Gregory M. Saunders and Jordan B. Pollack, (1996).
"The Evolution of Communication in Adaptive Agents"
In Technical Report GS-94-EVCOMM.Laboratory for Artificial Intelligence Research Department of Computer and Information Science The Ohio State University, USA

 

nah, not really.

If you ever can write an AI that can use an bow and some arrows without any knowledge of what it is nor how to use it then contact me.

The speech ability is not what the intelligence lies in, that is merely knowledge and identification.

 

Baal Zeb,
But you would have to admit that Chineese Room Argument is logically correct,isnt it?As you just mentioned in your above post The speech ability is not what the intelligence lies in, that is merely knowledge and identification.,i would say that Some fact grounding is required.Chineese Room argument,doesnt hamper the growth of AI.It infact,shows a path that must be followed for any further research comes into the subject.Probably, our current approach isnt exactly correct? may be we need something more,some more theoritical facts in AI which can guide us in a correct way to develop AI,something like Newton's formula...
IMHO.
just some thoughts...
bye!

 

Good analogy...Theoritical perspectives of AI can be considered as genetic science here...may they are in dire need of change...

bye!

 

yes, of course it needs to have everything translated to chineese in order to work. the goal-system could not be in english to function in chinesse for instance.
To that point i agree. But to give an ALF knowledge and grammar. That is not AI. Hope you agree with this atleast.

 

Yes.Very agreeeable.I hold the same opinion.As explained in this case of Chi Arg,you could have a translator program merely manipulating the strings to repond,which isnt exactly creating an AI.

bye!

 

If you start an AI with the ability to form any sentence, then you have gone to far.

A newborn child cannot talk.
Our AI is a newborn child. It cannot interact either but it uses a parameter based logic to solve these problems.

It of course uses references to learn grammar and new words, but atleast it has made those references by itself. If that is not AI, then we are niether intelligent.
Anything else is just a smart program.

 

The important thing i must say is,that a new born child relates his knowledge to the real world experiences and changes them accordingly,this gives weightage to concept of Neural Network...isnt it?

bye!

 

the visual neural net is trained by the sensory of touch.
The DB does not use NN, merely percentage data train the DB.
But the Visual NN and the other Sensories (such as Audio NN) train the DB.

does that make any sense at all?

When talking it relates the real world experience to the speech cortex and therefore evolves the ability to strengthen the empirical value in the entries.
If it is the other way around then it calculates the Fiscal Value and the trust for the one claiming it. If the ALF trusts him/her then it is probably true. Else it will add it as a entry in Economical Thinking and therefore add it to a list of items that needs investigation. The result from this investigation can raise or lower the trust (Fiscal Value) for a particullar person.

Not to fuzzy i hope.

 

I think the point of the Chinese room experiment is to illustrate that without a fundamental understanding of how intelligence in the mind works we cant ascertain whether a system is or is not intelligent. It demonstrates the short comings of the turing test. Testers can be fooled.
It also demonstrates that semantic understanding does not have to exist in a system for its to simulate a mind, yet postulates that to be a true AI, the system must have semantic understanding. This shows again that our understanding of the fundamental properties of intelligence is lacking.
I subscribe to the Strong AI philosophy and if a systems can demonstrate intelligent behavior then the system must have a semantic representation of its knowledge even though we cant explain how.
The Chinese room experiment is more an argument for and against sentience then AI and should be keep out of the AI arena.

 

Here is what i think "Intelligence does not come in paper-back".
The Chinese Room argument suggests that obtaining "alien" knowledge might be "impossible".
I am saying that my AI can do it but it needs to have the parameters in chinese too and since i lack knowledge in that particullar language i am choosing to build a ALF that can learn English without any start-up information.
I am not talking about any low-level AI like Exholio Neogenesis provided a link for. Where ALF's develop a natural language between eachother. I am talking about English.
But that is the simple part. How to make it understand how to use a axe on a tree, a bow and arrow when hunting a deer, how to use a bucket to transport water. Those tasks are impossible for other AI and it is the basis of mine.
It is also the basis of human intelligence.

I am not born with the ability to speak. I do not have a database of words and i do not have grammar. Why should AI have to?

If anyone disagrees with this then i do not see a very brigth future for the human race and Einstein was right when he said :
"There are two things that are static. The first is that the universe is constantly expanding, the second is human stupidity. And i am not sure that the universe is expanding"

 

Your AI must have some form of syntax or all you would have is a blank source file. "Intelligence does not come from paper-back" your source file being written on modern paper (a computer), ergo your program has no intelligence in your view. This is also the view of the Chinese room, that "Syntax by itself is neither constitutive of nor sufficient for semantics". It states that because all computers are based on Turing machines and that Turing machines use formal syntax that we should be very careful what we call intelligent, even if it acts like a human.

Of course I don't follow that line of reasoning.

 

Baal Zebul,
You said :

Which as i understand is either Wrong,or you dont have grasp of concept.It never makes any suggestions of any sort like obtaining some outside knowledge like chineese might be impossible.

Here's what is Chineese Argument all about :

I dont know what are you trying to say?in all probabilities its different than what i have just mentioned.

One reason given for the difficult time that modern AI systems have in passing the Turing Test is that since these systems work with representation of reality in form of symbols,they often sidestep the issue of how to ground these symbols in reality.

Your program knows what is Abate in english and where to use it for replying and manipulation to make it look as if it was an intelligent reply,but does it know what exactly is Abate.Has it ever used Abate in real life to solve its problem etc...? that is where we become intelligent,grounding of facts to reality...
i hope i clarify my point...
Anyways a company to run for a 16 yr old can be too much dont you think?...

Please Register or Log in to view the hidden image!

bye!