Genesis - Oct 18, 2011:

May I ask you for your definition of “True AI”?

By “True AI” I mean that MindForth is truly a mind and is truly intelligent. It is not an Eliza-style “fake” artificial intelligence. It does not use statistics to handle natural language. Instead, the AI Mind conceptualizes the natual-language words and thinks by means of association from concept to concept.

Yesterday I programmed for hours on MindForth and I added a neat new feature. Now when MindForth encounters a new English noun in the NewConcept module, it saves the new word as a “cogpsi” variable and waits until there is a lull in human input. Then the AI sends the “cogpsi” concept into the WhatBe module and asks either “What is…” or “What are…” [the new thing(s)].

Thus far today in the morning I have worked on the JavaScript artificial intelligence and I have solved the upthread problem of the AI saying “ROBOTS REPAIR WHAT DO I REPAIR”. The default “I” ego-concept was creeping in because I had neglected to restrict the NounPhrase module to looking only for a subject, not a direct object.

When I have made some advances in my AI work, I have posted a report here in this thread. In recent weeks I have been working on removing vestigial, disruptive influences from the AI software, with a near-term goal of achieving a rock-solid functionality in the AI thinking. If the AI currently thinks a thought that is obviously erroneous, it is because a free-radical concept has commandeered the NLP software and is not associating properly to the knowledge-base engrams. As I tweak and refine the AI-Complete software, we may expect to see in both Forth and JavaScript the emergence of an intelligent, thinking person.

Arthur T Murray - Oct 18, 2011:

Genesis - Oct 18, 2011:

May I ask you for your definition of “True AI”?

By “True AI” I mean that MindForth is truly a mind and is truly intelligent. It is not an Eliza-style “fake” artificial intelligence. It does not use statistics to handle natural language. Instead, the AI Mind conceptualizes the natual-language words and thinks by means of association from concept to concept.

 

Thinking is not the definition of Artificial General Intelligence (or True AI), its just one of the components. You can’t have a component and say, aha, this is True AI.

This AI thinks and actually does many intelligent tasks, but I don’t see the creator claiming it to be true AI do I?

Your definition of “True AI” is quite different from the accepted definition people go by. Which is the ability to perform “general intelligent action”.

Your AI has no done that. That is why people do NOT call it “true AI” because it hasn’t passed the requirement for be a true AI by the most accepted definition. Do you understand that?

Definition of True AI is what the AI does, not what it “IS”! You can have a fully complete version and it still won’t do a thing. An AI becomes a “True Genuine AI when it can accomplish general intelligent tasks”.

Calling your system “true AI” doesn’t help or promote it, it actually hurts it as you have experienced. I know retracting your statements will be too hard for you after all these years but at this point and with the reputation you have garnered with everyone whether or not they have a computer science degrees or just a regular joe, downwards is the only way forwards.

Secondly, you need a web design for your site. No one is going to take you seriously when your information is scattered across the surface of the earth. Plus the fact that its a wall text. I doubt someone will spend more than 30 seconds on your site, they will feel life passing them by.

Actually organize your work so you can garner a little bit of credibility.
Look at others in the field: http://opencog.org/ http://www.numenta.com/htm-overview.php

I’m only saying this to help you and about organizing your work. Take a look at the PDF of Jeff Hawkins HTM.

Notice how organized and appealing it is? Its not a wall of text. And its very informative and detailed with nice drawings and illustration. Plus peusdocode.

There is not elaborate words that sperate the reader from the material. You read it once and you understand it whole-ly. While most after stomaching that fact they are about to read a wall of text. Come out understanding little about your system or none at all.

Stuff like “Sensorium” are candidates of expulsion if I were you. These words doesn’t make your AI anymore interesting, nor does it help your readers understand what’s actually going on.

Infact if you had your work in order, you won’t need to bombard every forum known to man. People would visit your site and check back anxiously for updates. And if your AI actually did anything, you would be a billionaire by now.

Do you know how much money big corporations like Google, apple, Microsoft would pay for a true Ai system?

Apple bought “Siri” which they incorporated into their i4s for 200 million dollars. LOL if your AI was worth anything, wouldn’t they be knocking at your door?

Genesis
Let me share with you (all) my ideas on this!

I fully coincide with your thought, there are too much people making “false announcements’ about their skills and creations, the truth is more obscure and should be far more modest, may be someone out there did something and don’t wanted to show it up until it’s fully functional, or at least passes a good AI test. ¿but what is an AI test?

This forum if so prone to “blah blah” (I sometimes.. also contribute ) that you cannot take all things seriously,

I have made some good intents on this forum in making a ‘sort of’ academic-intelligent idea exchange allowing brainstorming, and found some good answers, but also bad ones (somewhat aggressive). Even though some good intended people in this forum (including the moderator, the owner and some clever members) did give good coaching and discouraged them from being disrespectful.

After a while, by reading the posts, you learn (by yourself) who is who, who answers intelligent stuff like: Andrew Smith, Dave, Erwin, C R Hunt, Gary, and many others, and other people who are making Spam (sometimes very boring and noisy) like 8pla and Arthur, among other newbies.

After this, you will see also many who are wanting to use this forum to make free-commercial adds and posts, to enhance google’s find outs like myCyberTwin, and some one else from some sponsoring-or-not companies in the wild.

I personally came into this forum to learn, exchange ideas and give back some teaching, in my humble opinion. I am really making a chatbot engine, a dep one, using NLP and a hybrid approach, many of my system information is public, and many parts have been publicly disclosed, (they are not open sourced) I did a framework to allow people develop a chatbot easily, but stalled against many problems which I am addressing since 5 years from now, many problems are still unsolvable on the current state of the art, and I am contributing to find some solutions at least ‘seem-to’. My approach is dual: commercial and research (for my PhD) I don’t want to patent it but hate the idea of being copied, because no one pays my salary for this, so I need to get some bucks out of them. Anyways you can find my publications on my university page: http://web.fi.uba.ar/~ahohenda there are many good samples to read form there, most are in Spanish, I translate some of them from time to time.

I must say it here, I tried out many bots, perhaps too many, and got sometimes fooled by some of them on selected answers, which gave me some “very clever” answers! wow! I said!

But as a 30 year long programmer & researcher on AI and knowing automaton theory, I can test them in an eye blink to tell out if they use AIML or not, and most of them use this approach (I guess 99%)

Some others (a few) are twisted-AIML with some tricks as “stemming” to allow them to ‘handle’ highly inflected languages like Spanish (my specialization) in a less-than-decent way (English spoken people usually don’t understand the big and painful misses of this approach)

But no one of them dare understand a single nominal/verbal/pronominal/adjectival part of a sentence, do a some inherent calculation or math, manage sets of objects, reasoning even in simple way, or have memory and context other than selected variables to show up as intelligent, no one solves anaphoric and co-reference stuff, no one disambiguate even a little! They (all) mostly do not make text analysis (some use wordnet as a free-simple ontological database) to allow some smart-ass ‘thesauri’ searching widening up the possible matching of the matching system, being more spectacular, like ChatScript, and RiveScript did a good job!.

Having said this, my last ‘boring’ discovery was myCyberTwin, who came into this forum saying they had found the Holy Grail, I am sorry to say I tested the online chatbot, who is a sample of their whole system and I found nothing new, nothing smart and nothing robust at all, nothing but a simple AIML clown!

Hope this helped and enlightens you a little!

cheers!

Genesis - Oct 19, 2011:

Stuff like “Sensorium” are candidates of expulsion if I were you. These words doesn’t make your AI anymore interesting, nor does it help your readers understand what’s actually going on.

Gratters, Genisers! You have written a masterpiece of http://code.google.com/p/mindforth/wiki/MentifexBashing in a sputtering, apoplectic style reminiscent of Garry Trudeau, Mike Royko and Hunter S. Thompson.

> People would visit your site and check back anxiously for updates.

“Rem acu tetigisti!’ as they used to say in ancient Rome, “You have hit the nail on the head.” You would not believe the user logs of the Internet masses coming to fetch the very latest free AI source code. And I take my role as Mentifer bringer-of-mind very seriously.

The barista here at this Internet cafe just said to me:

She: “Are you doing anything special today, Arthur?”
ATM: “Oh, just changing the world by introducing a whole new species of artificially intelligent creatures.”
She (laughs): “And other than that, nothing?”

> Do you know how much money big corporations like Google, apple, Microsoft would pay for a true Ai system?

They don’t have to pay for Mentifex True AI. It is free at http://www.scn.org/~mentifex/AiMind.html

Arthur T Murray - Oct 19, 2011:

Gratters, Genisers! You have written a masterpiece of http://code.google.com/p/mindforth/wiki/MentifexBashing in a sputtering, apoplectic style reminiscent of Garry Trudeau, Mike Royko and Hunter S. Thompson.

I’m not bashing you. The fact remains that you need to organize and express your work alot better than you have if you actually want to be taken seriously. No one wants to go through this.

Its 2011 not the 90’s if you haven’t noticed…

But yet again you avoided the part of my post that really mattered. That part that talked about the actual definition of True AI and how your AI does not fit into that definition. You know it, I know it, we all know it.

You’re not delusional, you know very well the fine lines and the facts. You just choose to portray yourself system as such because you think it would help your system. But the fact is, it hasn’t but you can’t turn back now so you continue the same path you created.

You said that Ben with his “Opencog” is has a plan till 2020 to roll out a true AI and that the race just got thighter.

But the race didn’t get tighter, according to you. You already won. You have the first true AI. You solved AI. So what race is there to be?

Infact, lets look at Ben’s OPENCOG roadmap and compare it to what your “true AI”.

2011-2012
- “Integrated “toddler-level” intelligence for video game / virtual world characters.”
Your AI doesn’t have “toddler-level” intelligence, it doesn’t even do anything that HINTS of intelligence.

- “robustness, flexibility and adaptive, interactive learning “

Your AI does not learn what so ever but repeat regurgitated inputs over and over again.

- “Simple English language dialogue, e.g. answering questions about its environment and state, taking instructions, asking clarifying questions when the instructions aren’t understood, etc.”

Your AI can’t even make a cohesent or even a statement that hints of intelligence talkless of the above.

- “Integration with OpenBiomind for inference-based meta-learning – the ability to learn about the learning process.”

Your AI doesn’t learn in the first place so it has no capacity to even learn about the learning process.

- “Inferential analysis of spatio-temporal scenes, e.g. the ability to answer questions about the relationship between entities without directly observing them.”

Again your AI is millions of miles away from accomplishing anything like this. It has yet to give hint of intelligence.

- “Inferential analysis of spatio-temporal scenes, e.g. the ability to answer questions about the relationship between entities without directly observing them.”

Again, your AI is not even close.

So your AI can’t even do anything remotely close to whats listed above and yet its a “true AI”?

You are not fooling anyone dear.
> People would visit your site and check back anxiously for updates.

Arthur T Murray - Oct 19, 2011:

“Rem acu tetigisti!’ as they used to say in ancient Rome, “You have hit the nail on the head.” You would not believe the user logs of the Internet masses coming to fetch the very latest free AI source code. And I take my role as Mentifer bringer-of-mind very seriously.

Is that why you fill every forum known to mind with your AI theories? There is a reason why you were banned from all those sites.

Arthur T Murray - Oct 19, 2011:

They don’t have to pay for Mentifex True AI. It is free at http://www.scn.org/~mentifex/AiMind.html

That makes it even worse. You would think with a true AI and these companies knowing the gravity of the AI age, the fact that it will be a billion dollar industry are not capitalizing on it. You don’t see any company trying to integrate your “true AI” with their products do you?

In-fact if you paid them a million dollars, they will still refuse it.

Genesis - Oct 19, 2011:

I’m not bashing you. The fact remains that you need to organize and express your work alot better than you have if you actually want to be taken seriously. No one wants to go through this.

http://www.flickr.com/photos/kylemcdonald/2420003882/ is where a famous artist writes in 2008: “When I was in high school I was really into your theory of mind—it’s exciting that you implemented it last year!”

> Your AI doesn’t learn in the first place so it has no capacity to even learn about the learning process.

Au contraire, mon frere. Mentifex AI learns by asking questions. Just last night I ported one of the machine learning algorithms from MindForth into the http://www.scn.org/~mentifex/AiMind.html in JavaScript.

> You don’t see any company trying to integrate your “true AI” with their products do you?

I don’t know what the various ‘Net entities are doing with my AI programs in Forth and in JavaScript. People out there on the ‘Net realize that my attempts to create True AI are sincere and ambitious. Netizens understand that a True AI must start out as an extremely simple artificial Mind and slowly grow or “accrete” more and more AI functionality. Ben Goertzel is like the Albert Einstein of artificial general intelligence, but his OpenCog does not seem to have any functional demo like MindForth or AiMind.html available on the ‘Net.

By the way, the link you provided to http://home.earthlink.net/~gmayhak/M5_htm.htm (where the first “htm” is for “Hierarchical Temporal Memory”) looks very interesting and so I may soon embed that link as an AI programmer’s “Comment” in the http://www.scn.org/~mentifex/mindforth.txt AI source code, by way of “outreach” to other AI projects that I come across in my Web wanderings. Thanks for the link, and for your advice and comments!

Andres Hohendahl - Oct 19, 2011:

I personally came into this forum to learn, exchange ideas and give back some teaching, in my humble opinion. I am really making a chatbot engine, a dep one, using NLP and a hybrid approach, many of my system information is public, and many parts have been publicly disclosed, (they are not open sourced) I did a framework to allow people develop a chatbot easily, but stalled against many problems which I am addressing since 5 years from now, many problems are still unsolvable on the current state of the art, and I am contributing to find some solutions at least ‘seem-to’. My approach is dual: commercial and research (for my PhD) I don’t want to patent it but hate the idea of being copied, because no one pays my salary for this, so I need to get some bucks out of them. Anyways you can find my publications on my university page: http://web.fi.uba.ar/~ahohenda there are many good samples to read form there, most are in Spanish, I translate some of them from time to time.

I read your work on the page and your 8 challenges. Well atleast the ones that could be translated by Google. Very interesting to say the least and direct to some of the problems of NLP we have today. Will be cool if you can put them (your work) all together into one Intelligent system.

Andres Hohendahl - Oct 19, 2011:

I must say it here, I tried out many bots, perhaps too many, and got sometimes fooled by some of them on selected answers, which gave me some “very clever” answers! wow! I said!

But as a 30 year long programmer & researcher on AI and knowing automaton theory, I can test them in an eye blink to tell out if they use AIML or not, and most of them use this approach (I guess 99%)

Some others (a few) are twisted-AIML with some tricks as “stemming” to allow them to ‘handle’ highly inflected languages like Spanish (my specialization) in a less-than-decent way (English spoken people usually don’t understand the big and painful misses of this approach)

But no one of them dare understand a single nominal/verbal/pronominal/adjectival part of a sentence, do a some inherent calculation or math, manage sets of objects, reasoning even in simple way, or have memory and context other than selected variables to show up as intelligent, no one solves anaphoric and co-reference stuff, no one disambiguate even a little! They (all) mostly do not make text analysis (some use wordnet as a free-simple ontological database) to allow some smart-ass ‘thesauri’ searching widening up the possible matching of the matching system, being more spectacular, like ChatScript, and RiveScript did a good job!.

Well that’s one of the problem of humanity in large. We always want to follow in someone else’s foot step rather than to create for ourselves. So we end up with many variants of the same pre-existing product.

Andres Hohendahl - Oct 19, 2011:

Having said this, my last ‘boring’ discovery was myCyberTwin, who came into this forum saying they had found the Holy Grail, I am sorry to say I tested the online chatbot, who is a sample of their whole system and I found nothing new, nothing smart and nothing robust at all, nothing but a simple AIML clown!

Hope this helped and enlightens you a little!

cheers!

Well think about it… Its the only way they could actually get anyone interested in their work. Pretty good business strategy don’t you think?

I read your work on the page and your 8 challenges. Well at least the ones that could be translated by Google. Very interesting to say the least and direct to some of the problems of NLP we have today. Will be cool if you can put them (your work) all together into one Intelligent system.

I actually did this job, two years ago, my chatbot did talk already and answered over 8 million Spanish questions in 4 months, there were 3 initial deployments, but after this I got some commercial problems (the company who hired me, won’t pay), I simply stopped them from working, then the company did hidden a re-engineering of the scripts and migrated to another engine (a simple AIML-type) to go on with their clients and owe me since then for the service!

I retracted to go commercial for this time and are working hard to give the system a remote IDE framework to allow remote debugging and compilation. (this actually works fine)

Feel free to look at my last papers (on my site), the block diagrams are in English, and the text you can translate easily with google translate! from Spanish to English (every piece of software I write inside comments in English, but the manual and my country handles Spanish as well! (sorry)
A good (but commercial) description is here:
http://web.fi.uba.ar/~ahohenda/docs/AndyBot_Brochure.pdf
enjoy!

There has been some interest in porting the AI Mind to Lisp.
In comp.lang.lisp on Usenet, please see these messages.

http://groups.google.com/group/comp.lang.lisp/msg/b25f653f16b954dd

http://groups.google.com/group/comp.lang.lisp/msg/0e2fa92b14329c04

“Provocation or Food for thought? Web sites as AI Mind Persons”
is the title of an invited blogpost of mine published under the permalink

http://www.forum.santini.se/2011/11/provocation-or-food-for-thought-web-sites-as-ai-mind-persons

at the WebGenre Blog of Marina Santini. Angry comments belong
neither here nor there, but constructive comments are welcome.

Meanwhile Yours Truly Arthur has beeen trying to covert
the one and only True AI Mind into German at

http://code.google.com/p/mindforth/wiki/DeKi

with a mind-diagram of the bootstrap module at

http://code.google.com/p/mindforth/wiki/DeBoot

and with a listing of initial German vocabulary at

http://code.google.com/p/mindforth/wiki/WortSchatz

and now I want to thank everbody for their refreshing
friendliness and civility here at ChatBots.Org.

Arthur

http://www.scn.org/~mentifex/Dushka.html

is the beginning of an AI Mind in Russian called
Душка (little soul) which will be a Russian copy of

http://www.scn.org/~mentifex/AiMind.html

the JavaScript AI Mind in English.

Author, I have read this entire thread on AiMind and have visited your website, inspected your code and relationships between your “modules” and I still do not have a clear concept of what your actually attempting to accomplish?

If you claim self-awareness of your “system” by means of running a script that updates it’s variable arrays by user input, I fail to understand this relationship. Understand. I am not putting down your project by any means, but as someone that is in the process of building their own bot that uses JavaScript almost exclusively, I am aware of its limitations without a static database and server interface to the client scripting.

Am I missing something here?
Is there a more up to date version to look at?

Thanks.

Laura Patterson - Dec 14, 2011:

[...] If you claim self-awareness of your “system” by means of running a script that updates it’s variable arrays by user input, I fail to understand this relationship. [...]

I am claiming “thinking” now and “self-awareness” in the future. The modules relate to one another by serving as the mechanisms of thought.

> Understand. I am not putting down your project by any means, but as someone
> that is in the process of building their own bot that uses JavaScript almost exclusively,
> I am aware of its limitations without a static database and server interface
> to the client scripting.

There is no limitation imposed by the lack of “a static database”, because
the user input creates a dynamic database. There is also no need for a
“server interface to the client scripting”, because the thinking AI Mind
is entirely contained in the client-side JavaScript AI program that
comes to life in the MSIE browser.

> Am I missing something here?

You may be missing a freedom of detachment to view things in a new way.

> Is there a more up to date version to look at?

http://www.scn.org/~mentifex/AiMind.html is the most up-to-date,
roughly on a par with the more-difficult-to-run Forth version at
http://www.scn.org/~mentifex/mindforth.txt

> Thanks.

And thank you, fellow JavaScripter, for looking at the AI Mind.
Here I would like to say that a month ago I started to port my
English-thinking AI into German, and then I realized that I
should create the AI simultanously in Russian, because
the effort to create each module in German means not
much more effort to create a similar module in Russian.

There is also some new functionality that I plan to code
into the AI Minds in English, German and Russian,
namely a right-adjusted buffer mechanism for holding
onto words in natural language and slicing off their
morphemic endings to yield the basic word-stem.

- Arthur (Артур Артурович)

Andres Hohendahl

Su trabajo es muy interessant. yo vou ler todo.
Mi Español no es muy grand pero yo consigo ler.
Mi lengua nativa es el português, entonces hay muchas similitudes.

Genesis

Hope you are doing fine in your project. Just always keep in mind what I’ve told you by e-mail.

Arthur T Murray

I am not sure if I understand your bot’s approach. I’ve spent some 2 hours trying to put inputs and your bot keeps just saying I, I, I. There wasn’t a moment it could say anything more than I .. not very interesting… sorry.

Author,

Before your script can execute, you need to correct the following error:

fact.style.setAttribute is not a function
[Break On This Error] fact.style.setAttribute(‘position’, ‘absolute’);
AiMind.html (line 3817)

Beyond this I am not sure what other errors may exist since this attribute is not defined and causing your script to crash.

Hope this helps.