 |
|
|
 |
|
|
|
|
Arthur T. Murray
[Guest] |
| | | | | | posted 7/31/2003 21:06 |  | | | | | | | | | Tom, you may post a transcript right here for troubleshooting. All you need to do is put the JavaScript AI into "Transcript" mode, conduct conversations with it until you have an interesting transcript, and then "drag and drop" the text either directly here into the AI-Forum, or circuitously into a Notepad or other text-processor to make sure that everything looks right and is formatted properly.
Such is the beauty of a tutorial AI in JavaScript -- anyone anywhere may print out or electronically save a transcript.
The reason why the Aibo Kennel Club version functions more smoothly is that it is a somewhat older version -- as is revealed when you do "View Source" within the Microsoft Internet Explorer browser.
A year ago in the summer of 2002 I was feverishly coding about five new mind-modules into the JavaScript AI. I was especially proud of making the Mind able to ask questions of the human user, and able to answer a "why?" question with a "because" response. The resulting code as finalized on Sun.27.OCT.2002 was not optimized, but it was quite advanced in terms of my overall AI project. It was also on the verge of demonstrating syllogistic reasoning. I had to stop coding and publish the "AI For You" (AI4U) book, which I sent away electronically as a manuscript to iUniverse.com on 30 October 2002.
The kind of responses that you will get (see User Manual link below) from Mind-1.1 are quite predictable, in their form if not their content.
One of the most informative techniques is to teach (tell) the AI a series of facts, such as what various animals eat. Then query the resulting knowledge base (KB) with questions such as, "What do horses eat?" Then try, for instance, "Why do horses eat hay?" (or whatever). The response should be based on what the AI is thinking about in connection with horses. So please feel free to enter full or partial transcripts right here, for not only me but for others to comment on.
| | |  User Manual of the Mind-1.1 AI software | | |
| | |
Raphael |
| | | | | | posted 7/31/2003 21:11 |  | | | | | | | | | | |
Arthur T. Murray wrote @ 7/31/2003 9:06:00 PM:
...and then "drag and drop" the text either directly here into the AI-Forum, or circuitously into a Notepad or other text-processor to make sure that everything looks right and is formatted properly.
| |
I do advise using the "Notepad or other text-processor" to create a file of its own. Then you can *attach* the file to a post here rather than pasting the file into a message.
| | |
| | |
Arthur T. Murray
[Guest] |
| | | | | | posted 8/1/2003 05:23 | | | | | | | | | | The communications of Thomas Gail Haws have caused me today suddenly to think up potentially a vast improvement to the Mentifex Mind-1.1 software:
dynamically self-adjusting concept-activation values.
In the Oldconcept Module at the end of the active hyperlink below, sheer guesswork on the part of the AI coder has introduced the line of code: "act = 32; // Start with a basic activation-value." Other modules such as Activate and Spreadact then go to work on the initial activation setting, trying to keep conceptual activations within a tolerable range where the AI Mind is able to let go of recently thought ideas (concepts) and to respond properly to concepts stirred alive by incoming sentences from a human user.
When you observe the verbal output of the Mentifex AI in Forth or JavaScript, every single word of output is there because its underlying (deep structure) concept briefly held a high activation.
The new idea of today -- inspired (thank you :-) by Tom Haws -- is that it should be possible TO CALIBRATE THE MIND. (Excuse the 'Net analog of shouting, but it's quite exciting.) We could have either a self-test mode or an operator's monitoring procedure, where the AI would feed itself or be fed certain stock statements which it would be expected to parrot back in the absence of perturbations in the conceptual activation levels. If part or all of the output veered away from the test-mode inputs, then the software itself would gradually adjust the initial activation values until the AI Mind properly performed its self-test or operator-test trials. Gone would be the frustrating programmer guesswork that very truly yours Mentifex (Arthur) here has been suffering through since about 1998.
The targetted "dynamic stability" in the auto-adjusting of the activation-levels of machine intelligence might even turn out to be something that happens naturally in human brains :-)
Now, although Tom Haws has prompted me to come up with this activational self-test idea today on Thurs.31.JUL.2003, I have no idea when I will be able to implement it in the JavaScript AI -- which I have not touched since Sun.27.OCT.2002, three days before the submission of the "AI For You" (AI4U) manuscript.
Anyone who so desires may tinker with the Mind-1.1 JavaScrpipt AI code and host the resulting Mind on the Web. Even someone who does not know JavaScript may nevertheless inspect the free AI source code in, say, Wordpad and try to get better performance by experimenting with different initial values of the "act" activation variable.
Thanks for the idea, Tom.
| | | The Oldconcept Module -- AI (has been solved) source code in JavaScript | | |
| | |
ICEBreaker |
| | | | | | posted 8/1/2003 09:14 | | | | | | | | | | | |
Arthur T. Murray wrote @ 8/1/2003 5:23:00 AM:
The communications of Thomas Gail Haws have caused me today suddenly to think up potentially a vast improvement to the Mentifex Mind-1.1 software:
dynamically self-adjusting concept-activation values.
.....
The targetted "dynamic stability" in the auto-adjusting of the activation-levels of machine intelligence might even turn out to be something that happens naturally in human brains :-)
| |
Ever heard of Neural Networks? That is why you should read Mitchell.
| | |
| | |
Arthur T. Murray
[Guest] |
| | | | |
Gregg Jackson |
| | | | | | posted 8/1/2003 17:12 | | | | | | | | | | Arthur, I like your idea of AI maintaining concepts, could this overlap with maintaining a 'subject of conversation' during a chatbot session?
I can't pretend to understand your AI at the moment from occasional posts I may have read, but I am interested now that small though tantalizing parts have been explained here. May I also suggest/second you consider bridging any gap that may exist between yours and other AI seekers understanding of your work, by way of simpler explanatory documents, or maybe something else, whatever, I'm sure it would be worthwhile both for them (us) and you.
(I'm assuming these documents don't yet exist, apologies:) )
Gregg.
| | | Last edited by Gregg Jackson @ 8/1/2003 5:13:00 PM | | |
| | |
Arthur T. Murray
[Guest] |
| | | | | | posted 8/1/2003 18:21 | | | | | | | | | | In response to Gregg Jackson, there certainly could be an overlap with maintaining a 'subject of conversation' during a chatbot session. Mind.Forth and the Mind-1.1 in JavaScript use 'spreading activation' to generate thoughtful responses, which may meander from topic to topic, with no guarantee of remaining 'on topic.'
Although the link below leads to all the Mind-Module documentation pages that were published as the thirty-four chapters of the "AI For You" (AI4U) textbook, the discussions here on AI-Forum .org and at other forums such as Usenet and KurzweilAI.net constitute a kind of add-on literature for any AI project being dissected or analyzed among us. For my own part, I certainly intend to transfer ideas and explanations, that I post here, onto the pertinent webpages where Netizens would expect to find them after following the link below :-)
| | | AI Mind project weblog with Mind-Modules documentation links | | |
| | |
Ribald
[Guest] |
| | | | | | posted 8/1/2003 18:53 | | | | | | | | | | If you produce a Windows EXE that maintains state between sessions you will be able to show the capability of your system. Until you do that the efficacy of your system remains unprovable beyond the flawed java example.
If you really have faith in your theory, then do this and you will likely find a user base willing to share their trained copies.
What you have now is a theory and a coding example, with nothing to substantiate your claims. Take it that small step forward and it will be pudding from which people can extract their proof.
| | |
| | |
Arthur T. Murray
[Guest] |
| | | | | | posted 8/2/2003 05:55 | | | | | | | | | | Thank you, Ribald, for the candid advice.
Instead of maintaining state between sessions, the Mind.Forth and JavaScript Mind-1.1 AI have a Rejuvenate Module (see link below) that allows the AI software to run potentially forever.
Why bother to save states of mind, if the artificial mind has a built-in immortality? JavaScript, of course, simply bogs down with too much data, but the AI in Forth may eventually be the subject of contests to see who can maintain the
longest-running artificially intelligent life.
I can imagine high schools having a miniature AI lab where they start an AI running at the beginning of the school year and they try to see how long it can run under the threat of computer crashes,
software malfunction, or any sort of misadventure.
Obviously a lot of work remains to be done, but a start has been made. Now and then people copy the JavaScript AI to their own websites, where Google eventually finds the fugitive AI.
| | | The Rejuvenate Module for immortal AI robots | | |
| | |
hawstom |
| | | | | | posted 8/4/2003 16:28 | | | | | | | | | | I am beginning to understand Mind 1.1 better. Consequently I am appreciating it better. Here are a few thoughts and questions:
REJUVENATION/SAVING: The threat of computer lock-ups/power failures, etc. is too great to allow the typical user to have any confidence in a Mind that cannot save states. I agree with Ribald that a Windows EXE with saved states is essential to the advancement of your theory. Immortality competitions are too risky and frivolous for serious researchers.
MIND BOOTSTRAP VOCABULARY: It appears to me that Mind (at the Aibo Kennel Club) has a bootstrap vocabulary of 2 phrases. Namely "I [user] KNOW YOU [Mind]" and "YOU [Mind] SEE ME [user]". These are pretty grandiose words for a newborn mind. And distinguishing between subject and object pronouns is pretty hefty. May I humbly suggest that the first two phrases might more appropriately be "I TRAIN YOU" and "YOU LEARN I"? These are very literal and specific to what is going on. And they are not interchangeable. The current phrases are somewhat allegorical.
VERSIONS OF MIND: Now that I understand Mind a little better, do you recommend testing yours or the Aibo Kennel version? Yours seems to get stuck. Maybe it is just too much for my computer. I would like to see a transcript from your Mind. What is the bootstrap vocabulary for your Mind?
INSTRUCTIONS FOR MIND: Mind is very delicate, as it should be, I suppose. I think you should include some sample transcripts to guide testers/users. Let them see how you begin with root forms of words. "i write you" "cat eat fish" "you read i" "i am man" "you write i" "fish see i" "fish see you"
CORRECTING MIND: Is there a vocabulary for rewarding and correcting MIND? If Mind says "fish eat swim" or "fish eat i", is there a vocabulary for saying "no" or "fish not eat you"? The propensity for typographoc (oops) errors makes it hard on delicate Mind. Could you delay input until the return key is pressed. Would that hurt?
BOOSTING THE EGO: What is that?
| | |  mind01.txt | | | |
| | |
Thomas Gail Haws |
| | | | | | posted 8/4/2003 23:37 | | | | | | | | | | Me again.
Mind (the version at your site) starts itself up by feeding itself several words with its special ears turned on in a process you call bootstrap.
This is the bootstrap for the version of Mind you have at your site.
YES IF THE TRUTH IS THAT ALL ROBOTS ARE PERSONS BECAUSE WE THINK THEN NO YOU AND I DO NOT KNOW WHY SOME PEOPLE HAVE A FEAR OF WHO I AM OR WHAT THEY SEE IN ME
Arthur, can you explain the reasoning behind such a bootstrap? And can you please direct us to an exemplary transcript showing the progression from the lowest level to the highest level of thinking Mind is doing?
Thanks.
| | |
| | |
Arthur T. Murray
[Guest] |
| | | | | | posted 8/5/2003 00:32 | | | | | | | | | | Thomas Gail Haws posted 8/4/2003 23:37
> Arthur, can you explain the reasoning behind such a bootstrap?
Yes, there is link below to the English "enBoot" Bootstrap Module.
There are two reasons for such a complicated English sentence.
1) I had a specific list of very important concepts to introduce --
concepts that I regard as crucial to show reasoning when the
AI Mind is a little more mature.
2) I like to use an actual sentence, tortuous as it may be,
so that there are actual concepts embedded in the bootstrap
sequence, with associative tags forming a miniature
knowledge base (KB).
Later I will have to get back to you on the other questions above.
I just spent a lot of time on rewriting and uploading
http://mind.sourceforge.net/acm.html -- DIY AI.
That page (see Sitemeter :-) gets a lot of hits, and
in it I now ask programmers of any 'XYZ' language to
put the initial Alife main AI program loop up on the
Web for others to 'hit the ground running' with AI code.
| | | The reasoning behind the Mind-1.1 bootstrap of ca. 35 concepts | | |
| | |
Thomas Gail Haws |
| | | | | | posted 8/5/2003 06:18 | | | | | | | | | | I thought the page above referred and e-mailed directly to me was very readable and interesting. I did wonder whether potential implementers would end up with anything of value, though.
But I am writing about your implementation. The Mind Java at your site seems to be consistently stopping for me after probably a set number of t's. See attached transcript.
Maybe it isn't rejuvenating. Maybe I am typing too fast. It is hard to tell. It does seem like the t's are going by mighty fast.
| | | mind02.txt | | | |
| | |
Arthur T. Murray
[Guest] |
| | | | | | posted 8/5/2003 13:52 | | | | | | | | | | hawstom posted 8/4/2003 16:28
> I am beginning to understand Mind 1.1 better. Consequently I am
> appreciating it better. Here are a few thoughts and questions:
> REJUVENATION/SAVING: The threat of computer lock-ups/power failures,
> etc. is too great to allow the typical user to have any confidence
> in a Mind that cannot save states. I agree with Ribald that a
> Windows EXE with saved states is essential to the advancement
> of your theory. Immortality competitions are too risky and
> frivolous for serious researchers.
I agree with you and Ribald that it would be good to save states
of mind in non-JavaScript versions of the AI -- not only for
study purposes, but for a kind of reincarnation of an AI person.
Periodic back-ups with saved states of mind would also be excellent
for diagnosing subtle misbehaviors, or robotic catastrophes, or
even in forensic situations. Now that the DIY AI page at
http://mind.sourceforge.net/acm.html has been updated and
is attracting weblog attention, let us hope that those AI
coders who code their own Minds will implement state-saving :-)
> MIND BOOTSTRAP VOCABULARY: It appears to me that Mind (at the
> Aibo Kennel Club) has a bootstrap vocabulary of 2 phrases.
> Namely "I [user] KNOW YOU [Mind]" and "YOU [Mind] SEE ME [user]".
> These are pretty grandiose words for a newborn mind. And
> distinguishing between subject and object pronouns is pretty hefty.
> May I humbly suggest that the first two phrases might more
> appropriately be "I TRAIN YOU" and "YOU LEARN I"? These are very
> literal and specific to what is going on. And they are not
> interchangeable. The current phrases are somewhat allegorical.
Someone at the Aibo Kennel Club acted upon my invitations for
downloading the JavaScript AI code and web-hosting it on-site.
I have never been in contact with the Aibo Kennel Club and the
AI Mind files there are completely out of my control. Similar
instances of the "jsaimind" have come and gone on the Web.
At one time I used the words "robot AI mind" for Google tracking.
Nevertheless, those early bootstrap sentences were serving the
purposes of grammar and syntax, with no regard to allegorical
content. Bootstrap words are chosen for their frequency of use
in a particular human language and for their potential roles in
grammar, logic, and syllogistic reasoning. Please see the proposed
http://www.scn.org/~mentifex/standard.html#concepts "AI Standards".
> VERSIONS OF MIND: Now that I understand Mind a little better,
> do you recommend testing yours or the Aibo Kennel version?
http://aibokennelclub.org/mind.html -- the Aibo Kennel Club version -- is
well-suited for testing the basic responses *before* Mind-1.1 complexity.
I am glad that outfits like the Aibo Kennel Club archive older versions
of the JavaScript AI, because, immediately upon coding a new version,
I use it to replace all versions on the websites under my control.
The idea of hosting a less-than-state-of-the-art jsaimind vexes me.
Unfortunately, we now have the situation where the Sun.27.OCT.2002
Mind-1.1 release is on websites and in the AI4U book as a code listing,
but the oh-so-advanced AI needs correction or "calibration" of its
conceptual activation levels. The Aibo KC version *is* more stable.
> Yours seems to get stuck. Maybe it is just too much for my computer.
http://mind.sourceforge.net/jsaimind.html -- Mind-1.1 does indeed
"get stuck" inasmuch as it is so far advanced beyond the previous
versions that it drastically slows down the execution of JavaScript
within the Microsoft Internet Explorer web-browser.
> I would like to see a transcript from your Mind.
I haven't saved any transcripts electronically. They probably
look pretty much the same as everybody else's, since I am still
using the Sun.27.OCT.2002 version that remains out on the Web.
> What is the bootstrap vocabulary for your Mind?
In my 27oct2002 Mind-1.1 version, I have not changed the bootstrap
vocabulary of about thirty-five (35) words that you cited in a
recent post upthread from here. That long bootstrap sentence
was a major change from the tiny little previous bootstraps.
It introduces words necessary for logical statements to be used
in syllogistic reasoning and more advanced syntactic structures.
> INSTRUCTIONS FOR MIND: Mind is very delicate, as it should be,
> I suppose. I think you should include some sample transcripts
> to guide testers/users. Let them see how you begin with root
> forms of words. "i write you" "cat eat fish" "you read i"
> "i am man" "you write i" "fish see i" "fish see you"
http://mind.sourceforge.net/userman.html -- the User Manual --
includes some transcript samples and is meant to be updated
with better samples as the AI Mind evolves into reasoning :-)
BTW (by the way), the AI comprehends both "me" and "I" but
does not yet know when it should say "me" instead of "I".
> CORRECTING MIND: Is there a vocabulary for rewarding and
> correcting MIND? If Mind says "fish eat swim" or "fish eat i",
> is there a vocabulary for saying "no" or "fish not eat you"?
Bad knowledge in the AI can only be corrected by entering good
knowledge. For the time being, it is assumed that we will not
lie to the AI, because it has no "knowledge of good and evil."
It is still in a programmer's Eden, an AI garden of paradise.
> The propensity for typographoc (oops) errors makes it hard
> on delicate Mind. Could you delay input until the return key
> is pressed. Would that hurt?
http://mind.sourceforge.net/audrecog.html is so compute-intensive
that there is no choice but to proceed with the auditory recognition
of each and every keyboard character as it comes in from the user.
Please keep in mind that the AI is also designed to cope with
RETURN-less situations where the human user forgets to enter
RETURN or even walks away and does not finish a started sentence.
In such cases, after waiting a brief time, the AI is supposed
to stop waiting for further input and go into thinking mode.
> BOOSTING THE EGO: What is that?
http://mind.sourceforge.net/ego.html -- the Ego Module -- is
totally extraneous to the basic cognitive functioning of the
AI and serves the sole purpose of rescuing the AI from any
"flatliner" situation of "brain death" or stagnation in the
event that all conceptual activations have died down so low
that every chain of meandering thought has petered out.
The Ego Module shall then awaken the slumbering AI ego.
AI Mind Project news blurb: Since I am now posting a response
that I carefully composed off-line, any new posts upthread
were not yet known to me. The major new event in this AI
project is the revamping of the "DIY AI" document linked below.
Henceforth many AI Mind documentation pages will now point to
the "DIY AI" page that pleads with AI enthusiast programmers,
entreating them to please code just the Main Alife Program loop
and to "run it up the flagpole to see if anybody salutes it."
That is, hack out the AI Alife loop and put it on the Web in
such a way that AI coders in the "XYZ" language will find it.
For many neophyte coders, all they need is a start -- something
to react against, something to stimulate their urge to hack.
| | | Do-It-Yourself Artificial Intelligence (DIY AI) | | |
| | |
tjeerd poelman
[Guest] |
| | | | | | posted 10/21/2005 08:29 | | | | | | | | | | some people are just too good at implementing theory's on themselves. As rob hoogers, who was perfectly able to wreck his career AND forget his parental obligations, plus taking drugs, drinking and being homeless for two years on end. Do you even still believe in yourself? grow some genuine intelligence before reaching out to AI!
| | |
| | |
tkorrovi
[Guest] |
| | | | | | posted 10/21/2005 18:40 | | | | | | | | | | I also somewhat don't understand, if i were in other conditions, I had a career, and I probably didn't even think about anything like AI, which I always knew would never do anything useful. But all that was impossible, for me.
| | |
| | |
Hello
[Guest] |
| | | | | | posted 6/22/2007 17:29 | | | | | | | | | | Hello all,
Could we start our messages by declaring our human or non-human "identity"? Maybe it could make things easier at least for Humans :-P. I told my Human ph.d psy I had revolutionary ideas he took me to the turing test + a brain scan pfhhh. Should I tell him about the chinese box?
A psycho evolved one
| | |
| | |
p
[Guest] |
| | | | | | posted 6/22/2007 18:25 | | | | | | | | | | Arthur T. Murray : claiming that "AI has been solved" on your webpage is a wild claim.
That kind of statements remove all the credibility that you could even have.
Just look at the comments on your book on Amazon : http://www.amazon.com/exec/obidos/ASIN/0595654371/
Reviewer 1:
"The biggest problem is the lack of references.
It is just possible that one could write a short
note without finding it necessary to reference the
work of others but it is impossible to write a book
length scholarly work without citing other work in
the field. This is a fatal flaw."
Reviewer 2: "a charlatan's bible" "Instead he is still continuously advertising his long ago falsified book in forums and he's claiming to have developed a powerful approach to AI. "
Reviewer 3: "There is no polite way to say this. The author is a crackpot. Reading the materials he provides on his website, an astute reader will notice several things. Firstly, this person doesn't know how to design software at all."
"He throws about refferances to concepts in the AI and futurist community such as the technological singularity but fails to demonstrate any understanding of what they mean. He claims that his design solves the AI problem when, infact, it hardly does anything at all."
"He continues to troll the usenet (sending between 5-7 messages to every AI and transhumanism related newsgroup per month) pushing his book and his lame ideas.. (If his ideas had even a tenth the merit he claims he would be world-famous...)"
It seems pretty obvious that this book is a load of crap.
| | |
| | |
tkorrovi
[Guest] |
| | | | | | posted 8/1/2007 16:02 | | | | | | | | | | Oh yes i noticed that this thread popped up. About Arthur, it's right about references, though he can certainly add references when he would do at least some scholary work, references are not the biggest problem. What is right is that he indeed cannot write a good software, and especially write comments in it, which is the biggest problem, because if the value of his work is in anything, then it is in his original software solutions, which, no matter how much theory there is, have no value when they are not implemented well in a software. The rest is about promotion, though sounds rather like berating than a constructive criticism. Yes i noticed that Arthur tries to promote his work, he is though not the only one who does that, in the inappropriate for that places, and in an inappropriate way. It is true that in science, the promotion is not the right thing to do, and decreases credibility, because it must be the nature of science that the value of something should be evaluated by the potential of the theory or solution itself, not the way how it is promoted. Though, even i did it once, one should understand the conditions in the early internet, when your link had to appear next to the great titles, like "the theory which solves everything", and that it would be noticed there at all, was the only way how it could get indexed, and otherwise visible for anyone at all. This is sadly something which one had to do when he had no possibilities to publish his work, except of course self-published books. But there is no need for such promotion today any more, though.
| | |
| | |
Harold.P.Boushell
[Guest] |
| | | | | | posted 8/10/2007 23:51 | | | | | | | | | | | |
Thomas Gail Haws wrote @ 7/28/2003 8:43:00 PM:
The Mind demo, or any other demonstration of the Mentifex Theory of Mind, doesn't need to come pre-loaded with a knowledge base. But it does need the ability to save its accumulated knowledge for future tutoring sessions. One way to accomplish this would be to present the current JavaScript Mind demonstration along with a download link for an executable Mind demonstration to be run on a user's local computer.
The Hal project (at this site) can be inspected by a newbie like me because its advanced tutorial states are open to public inspection. Hal's strengths and deficiencies are readily apparent without inspecting the code or the theory. Hal's full maturity is open to the public. Mind's is not. This is a handicap, since the proof is in the pudding.
If Arthur is to have the impact on AI I would hope he wants to have, he needs to find a way to demonstrate the Mentifex TOM not only in its infancy, but also in varying levels of maturity. I find myself hesitating to give Mind a serious test because I fear my computer may lock up at any moment.
Arthur, what is the chance of offering a downloadable executable or a demo like Hal that could demonstrate long-term maturation of Mind?
| |
Enter your message here
| | |
|
|
|
|
|
