The genesis of a new generation. A generation of true artificial intelligence is on the horizon.
I’m working on a “true” AI system in the making. I’m not here to make claims or say I have a “true” AI system or tell you “I solved AI”. Far from that. But I will present my theory that I have been working on. There are three phrases to it, I will cover just two of them. I will also try update the thread regularly with updates. I also address the definition of “true” AI, what routes I’m taking to accomplish it, the major land mines in the field of AI and finally my three phase system.

True Artificial Intelligence

or what some call “Strong” AI. But what is it? Well most people’s definition is an AI system that is close to, matches or exceeds human intelligence. But that doesn’t bring us any closer to having a unified definition of AI, does it? We would have to define what Intelligence actually is.

What is Intelligence?

Intelligence is not information nor is it knowledge. We tend to brand theses together. Intelligence is the process of gathering and identifying information. Proof are the dozens of medical cases where people lose all their memory, with it goes their knowledge of everything. They are then required to learn everything over again like a child; how to talk, eat, walk, etc…

If intelligence was knowledge then they would have stopped being intelligent the moment they lost all their memory? surely not. Their ability to relearn everything from scratch successfully is where intelligence hides. Its the process of acquiring and identifying information/knowledge.

Human level intelligence is not Human level knowledge, but the means to acquire it. So therefore the determinant of a “true” AI system is not whether it has adequate human level knowledge, but whether or not it can acquire it. There in lies intelligence!

What components are required for an AI to be considered a “True” AI system?

- The ability to remember
- The ability to think by remembering
- The ability to reason by thinking
- The ability to plan by reasoning
- The ability to strategize by planing

Major Pitfall and land-mine in AGI research?

The implementation of Emotions and Consciousness, it has held the research of AI back because the majority of those who are reaching AGI are building an AI based on that component. Its not required at all for a true AI system. Its only slowing down the process, Its not a necessarily component for an AI system.

Ariadne - Just one more Genesis will do.

The AI i’m programming is called Ariadne. Its not a system that is dated to be finished 10 years from now or 5 years from now. I facepalm when I hear someone say, they have been working on an “true” AI concept for 10, 20, 30+ years. Why stick with something that doesn’t work. I believe that is the reason we do not have “True” AI today, because we stick to our guns. We lose touch with reality that keeps trying to tell us, if it doesn’t work, dump it. What’s the definition of insanity again? Doing the same thing and expecting a different result?

Ariadne is a system not based on neural nets or database systems like Cyc or any other main routes we have today. Information is not directly inserted into the core of the AI system.  Every single thing is learned. Its a sort-of a bottom up approach but not a evolutionary one because there is a solid base. I don’t believe in Evolution because there is no direct proof of it, so its insane if I tried to use it as a model to build an AI system.

The First Phase of Ariadne

is what I’m working on right now. Its projected to be complete by the end of the year. The obviously timeline varies.

Phase one is like a test of the AI core, its not really what the goal of the AI is.

Here is a Design concept of how a completed phase will operate: http://img35.imageshack.us/img35/517/cognisense1.png

When phase one is complete, Ariadne will be able to remember, think and reason. Phase one is based purely on text. Phase one intention is not for the AI to understand the meaning of words ( though I can’t say its not possible). In example, “ball” means a sphere object. But to understand the text initself.

“Pusillanimous”. You probably don’t know or might know what this word means. But you sure know its a word right? But how? You know what a phrase is when you see one, you know what a sentence is. You know about punctuations and when to capitalize a word. You know what a paragraph is and so on and so forth. How can an AI acquire this knowledge without it being plugged directly into it like we see in today’s chatbots?

“Howesedfaboutttyrlnowfrt?” <—- What does that say?

If I inputed that into every chatbot today, it will see it as rubbish. But you are able to recognize the intelligent words in there and discard the rubbish. These are some of the issues that require intelligence.

When phase one is completed, Ariadne will be able to:

- Recognize words
- Recognize sentences
- Recognize punctuation and capitalization.
- Recognize when a user is typing and stops typing to take a break.
- Recognize when a user mistypes a word.
- Recognize when a user deletes a letter in order to correctly spell the word.
- Makes word predictions as you type.
- Recognize patterns in words.

Phase two goal is more in tuned with the goal of the entire project. Ariadne is now a Personal Assistant which is able to accomplish any computer task for the user:

- Finally able to understand the meaning of things.
- Able to use computer programs such as word, web browser, Photoshop, visual studio, etc…
- Because she’s able to understand the meaning of things, she will be able to write essays for you, help you with that 10 page research paper, solve that annoying math problem you are stuck on, create elegant programs for you in whatever language (c, c++, java, php), design web sites, so on and so forth.

Phase two is projected for summer of 2012.
Till then, the world awaits for Genesis!

Grand Designs.

Toby Graves - Oct 10, 2011:

Grand Designs.

Heh. I nearly said Intelligent Design.

Indeed, optimistic start

But you mostly told us about abstract concepts of your system. Could you perhaps tell us more about details of how you plan to implement them? You got me curious

Phase one sounds a lot like what google already does. How much of this ability is learning through input and how much is directed by programmers, I cannot say, but I imagine given the breadth of searches google sees daily, it must include a large component of the former.

Also, please define (in concrete terms):

Finally able to understand the meaning of things.

Though I guarantee, no matter what definition you give, you will be hard-pressed to find those who agree with you. Alas that’s the trouble with indefinite terms like “understand the meaning”. We know it when we do it ourselves, but try to quantify it, and all sorts of hell break lose. Look through some other threads on this forum and you’ll see as much.

Lastly, statements like this,

I don’t believe in Evolution because there is no direct proof of it,

make me question your credibility in general. (Not in relation to whether evolutionary approaches to AI are sound, just the part I quoted above.)

Frankly you sound like someone who’s thought a lot about the end product you’d like to see, but hasn’t spent enough time on the rigors of NLP to begin to comprehend all the little pitfalls that amount to a monumental challenge. Universities have collaborations dedicated just to building incrementally better parsers. If you can’t see how this could absorb 10 or more years of someone’s life, you just haven’t emersed yourself in the problem enough.

I hope you continue with this project—soon enough you’ll see what I mean. And I look forward to updates as your plans turn to real code.

Mark tM - Oct 10, 2011:

Indeed, optimistic start

But you mostly told us about abstract concepts of your system. Could you perhaps tell us more about details of how you plan to implement them? You got me curious

Well here is a brief explanation of how the core works. The AI receives input in the form of text (as of right now). The input goes through rigorous series of pattern recognition functions. The processed input are then streamed to memory.

Activation

Objects and Concepts in memory activate based on a threshold of similarities with the data input.

Objects

Words, images, sound and sequential patterns can be objects. Each time an object is activated it gets stronger. Each objects can be related to another with a relational cord link. Each time an object is associated with another, the link between each other becomes stronger.

Two or more objects become concepts when their relational cord link are so close that they pass a relational thresh-hold. For example, the word object ‘cup’ and an image object of a cup can be represented as a single concept.

Concepts

Concepts are two or more objects. Objects and Concepts are re-organized in memory by the MO (Memory Organizer). Concepts that have traits in common are brought close together by the MO to form a cluster. For example, the concept of cup and water. Each time this instance occurs, they reorganized closer.

Example

When the memory comes in contact with an input of “Barack Obama”, the concepts related to that input is activated, like a picture of barack obama in the white house. IF the input is a motion picture (like someone drinking), the “drink” concept is activated. If the input is a person driving, the activated concept will be a sentence “the person is driving a car”.

So, if I understand that correctly, the core is a relational database, containing items and relations between items. Besides the fact that you can do this by mining wikipedia, or using a database which already did this (see this), how are you going to use this information to make the agent say something interesting?

Nice to see so many people working on ‘true’ AI.

C R Hunt - Oct 10, 2011:

Phase one sounds a lot like what google already does. How much of this ability is learning through input and how much is directed by programmers, I cannot say, but I imagine given the breadth of searches google sees daily, it must include a large component of the former.

Not quite. Google uses a database of searches and the most searched terms appear on the drop-down list as you type. Since “what is my ip” is one of the most searched terms. when you enter “what” in google, what is my ip is the first thing on the drop-down list.

If I entered “Howesedfaboutttyrlnowfrt?” it surely doesn’t understand what that says. Neither does it think or reason.

C R Hunt - Oct 10, 2011:

Also, please define (in concrete terms):

Though I guarantee, no matter what definition you give, you will be hard-pressed to find those who agree with you. Alas that’s the trouble with indefinite terms like “understand the meaning”. We know it when we do it ourselves, but try to quantify it, and all sorts of hell break lose. Look through some other threads on this forum and you’ll see as much.

NATURAL LANGUAGE PROCESSING
In the dictionary, words have definition based on sentences with multiple words. For example: drink reads “to take water or other liquid into the mouth and swallow it”

In my AI, sentences are not stored in the memory (only sometimes), but the meaning of the sentence is rather stored. When you think about the word “drink” (a verb) you envision in your mind a picture or a motion picture of someone taking a drink. Well, that is the real meaning. A simulated pattern of the action of drinking is what is activated in the memory.

C R Hunt - Oct 10, 2011:

Lastly, statements like this, make me question your credibility in general. (Not in relation to whether evolutionary approaches to AI are sound, just the part I quoted above.)

Frankly you sound like someone who’s thought a lot about the end product you’d like to see, but hasn’t spent enough time on the rigors of NLP to begin to comprehend all the little pitfalls that amount to a monumental challenge. Universities have collaborations dedicated just to building incrementally better parsers. If you can’t see how this could absorb 10 or more years of someone’s life, you just haven’t emersed yourself in the problem enough.

I hope you continue with this project—soon enough you’ll see what I mean. And I look forward to updates as your plans turn to real code.

I stand wholly by my statement. Having read lots of writ-tens on AI. I gringe every time I read the association of evolution, natural selection, evo-biology, nature, or w/e with AI “This is how evolution did it…etc” Its quite annoying when you know those thing ain’t true. There isn’t one shred of direct scientific evidence that evolution (all life evolving from a single cell molecule) ever occurred, is occurring or will ever occur.

So when I see statements all the time about evolution in AI books, papers, theories, videos. It leaves a bitter taste in my mouth. But I don’t want to get into it because its in the realm of atheistic and religious debates which I don’t mind having in other forums where the platform welcomes it.

But here, lets deal with the AI aspect.

Here’s hoping that your one true AI will finally be able to settle the question of which is the one true god, given that there are currently so many to choose from. Also I expect it’s no coincidence that so many people believe the world will end next year, right about the time that you expect to be done creating your one true AI.

Mark tM - Oct 10, 2011:

So, if I understand that correctly, the core is a relational database, containing items and relations between items. Besides the fact that you can do this by mining wikipedia, or using a database which already did this (see this), how are you going to use this information to make the agent say something interesting?

Nope, the definition of database is “an organized collection of data”. I would be selling my core short if i said that is all it does. And my memory system is nothing like YAGO2. Plus there are different aspects of the systems not just the memory. There is the pattern re-cog, the thinking and reasoning module. So no, its not just a database. The memory is a organized collection of data, but that’s it. How the data is mapped and activated is the job of the other functions of the system.

The AI (for phase one) consists of Pattern re-cog module, memory module, thinking module and reasoning module.

(and of-course other hidden modules that are classified.)

Andrew Smith - Oct 10, 2011:

Here’s hoping that your one true AI will finally be able to settle the question of which is the one true god, given that there are currently so many to choose from. Also I expect it’s no coincidence that so many people believe the world will end next year, right about the time that you expect to be done creating your one true AI.

Well first of it, its not a “true AI” yet. Its the goal of the project to attain that status. I am no where close at the moment. I’m about 20-25% into phase one.

Secondly, this is a perfect example of confusing human level intelligence with human level knowledge. A general AI system is not determined by whether or not it has adequate human level knowledge or higher, which is what you are implying with your “god” comment.

We always have the idea that if an AI is to be a “true AI” system, it must do this or do that. Like find cure for cancer, etc… But that couldn’t be any more further from the truth.

An example of a true AI system is this:

We have chess AIs, checkers AIs, Spades, Hearts, Tonk AIs, you name it we have it. But we don’t have an AI that can play chess, checkers, spades, hearts, tonk, tic-tac-toe, backgammon, etc on the fly! What ever board/card game you present it with. Why? mainly because we don’t have AIs that can actually “understand” the game but rather AIs that work with predefined rules.

Uh oh, it looks like someone else has beaten you to it.

http://www.chatbots.org/ai_zone/viewreply/7108/

Personally I’m looking forward to the one true AI demonstrating once and for all the Earth is flat, since there is no direct evidence to the contrary.

Andrew Smith - Oct 10, 2011:

Uh oh, it looks like someone else has beaten you to it.

http://www.chatbots.org/ai_zone/viewreply/7108/

Personally I’m looking forward to the one true AI demonstrating once and for all the Earth is flat, since there is no direct evidence to the contrary.

lol… the difference between me and Authur is that he has not provided one shred of intelligent demonstration of his supposed “Aimind”. But if my AI does attain general AI, then I should be able to demonstrate it and not just come here to spew bs claims after claims.

For example, the “cogni-sense” program is a demonstration of the ability of phase one that touches all its capability and any claim that I make. You would would be able to test its abilities by connecting to an IRC channel and inputing text to it.

There will be also be a demonstration for phase two that also addresses all the capabilities that I may claim. (A simple ‘Intelligent’ Operating System desktop environment) on the web that will be made available to everyone.

A word of advice: If I make any claims and don’t demonstrate it, then disregard it and forever classify me as a looney. I can’t even assure you that my theory works till I’m able to complete phase one.

Authur on the other hand makes claims with no substantial demonstration. His Ai-mind that he has on the web doesn’t even do anything.

EDIT: Idk what you are inserting religious rhetoric into your posts. But first of all, get it straight, the bible said the earth was a sphere long 5,000 years ago, while the rest of the civilization believed its was flat.

Not quite. Google uses a database of searches and the most searched terms appear on the drop-down list as you type. Since “what is my ip” is one of the most searched terms. when you enter “what” in google, what is my ip is the first thing on the drop-down list.

What Google is doing is clustering words according to how they have been used most often, guessing about your input based on its experiences. This seems rather similar to what you intend to develop. Heck, even the image search associates text clustered near images with the image itself. That’s as close as you can get to what you call concepts without being able to really “see” images. (And they can do that too now.) How are you going to teach your system to process images and recognize objects?

If I entered “Howesedfaboutttyrlnowfrt?” it surely doesn’t understand what that says. Neither does it think or reason.

You’re right, it didn’t get that one. But I just tried “hjowsaboutknow” and it asked if I meant “how about now”, so I call that pretty good. And your second sentence is the exact reason why no one will believe that your bot is “True AI”, even if it does exactly what you’re proposing. One can always pull abstract words (“think”, “reason”) out of one’s hat and claim that a computer ain’t doing it.

I stand wholly by my statement. Having read lots of writ-tens on AI. I gringe every time I read the association of evolution, natural selection, evo-biology, nature, or w/e with AI “This is how evolution did it…etc” Its quite annoying when you know those thing ain’t true. There isn’t one shred of direct scientific evidence that evolution (all life evolving from a single cell molecule) ever occurred, is occurring or will ever occur.

I think what interests AI researchers is the idea of starting with a system that can be altered, refined, and mutated in response to external stimuli. This type of evolution has been witnessed first hand, both in the lab and in the field. Nobody’s talking about building an AI from a handful of organic molecules…

But I don’t want to get into it because its in the realm of atheistic and religious debates which I don’t mind having in other forums where the platform welcomes it.

But here, lets deal with the AI aspect.

I agree, I have no intention to discuss religion or atheism here, personally or in general. But talking about evolution is talking about observable science. And a science that has been applied to AI research (whether appropriately or not, which is a separate issue).

We have chess AIs, checkers AIs, Spades, Hearts, Tonk AIs, you name it we have it. But we don’t have an AI that can play chess, checkers, spades, hearts, tonk, tic-tac-toe, backgammon, etc on the fly! What ever board/card game you present it with. Why? mainly because we don’t have AIs that can actually “understand” the game but rather AIs that work with predefined rules.

Well, there was this group that taught a computer Civilization II by giving it the manual to read.

—-

@Andrew: Dang it, you’re cracking me up but stop trolling before Dave yells at us.

@Genesis: You’ve responded to Andrew’s posts with a level of professionalism that does you credit.* I’m looking forward to demonstrations of your system as you develop it.

*However, treating translations of the Bible as original text bothers the heck out of me. Consider this comment related to historical literature rather than religious use of said literature: The Bible said the earth was a circle, not a sphere. The Hebrew word “sphere” does show up in the Old Testament, but never in relation to the earth, which is always referred to as a circle. Okay. Enough Bible stuff. Let’s talk thinking machines!

Genesis - Oct 10, 2011:

Andrew Smith - Oct 10, 2011:

Uh oh, it looks like someone else has beaten you to it.

http://www.chatbots.org/ai_zone/viewreply/7108/

Personally I’m looking forward to the one true AI demonstrating once and for all the Earth is flat, since there is no direct evidence to the contrary.

lol… the difference between me and Authur is that he has not provided one shred of intelligent demonstration of his supposed “Aimind”. But if my AI does attain general AI, then I should be able to demonstrate it and not just come here to spew bs claims after claims.

I was waiting for this thread to get Aurthured, and I was not disappoint.

Genesis - Oct 10, 2011:

EDIT: Idk what you are inserting religious rederick into your posts. But first of all, get it straight, the bible said the earth was a sphere long 5,000 years ago, while the rest of the civilization believed its was flat.

I will be truly interested when, or if, there is some tangible demonstration of ANY of the concepts you are cogitating on here.  Until then it is just another vapor thread with lots of “concepts” (which are interesting, but as yet not demonstratively functional).

btw- IF an AI was ever convinced the Bible was actually “the word of God” and not just the writings of men remembered from the words of other men (and not even any women!), then it will be a sad day indeed for that poor gullible AI.

bbtw- Oh, and so the writings of Men can be taken in the proper perspective, 5000 Earth yrs is only “long” in human terms… the Universe, as we understand it, is in the range of ~14 billion Earth yrs old (or 16,000,000,000 times longer than 5000 yrs).