Thunder Walk - Jun 29, 2011:

Not to argue, but for the sake of discussion…

Discussion is healthy, and why we’re all here, I’m sure. Heck, even a bit of arguing, if done in moderation, is good!

Thunder Walk - Jun 29, 2011:

I’m confident that someone, somewhere, is working on everything to improve, expand, and modernize it.{AIML}

That very topic has been discussed here, as I’m sure you’re aware.

Thunder Walk - Jun 29, 2011:

I disagree that people continue to use AIML simply because they’re lazy or comfortable.

On this, we’re in complete agreement. And please don’t think that just because there is something “newer”, that it’s automatically assumed to be “better”.

Thunder Walk - Jun 29, 2011:

However, if only futuristic thinking, and a devotion to whatever is next is the soul purpose of this forum, and only commentary regarding yet another… and another… and another Proof of Concept bot is welcome (or understood) then, Steve was correct.  The stigma regarding AIML seems to exist, and I’m in the wrong place.

While a large part of the drive and focus of this community is the exploration of new technologies and/or new methods that might/could advance the field of Artificial Intelligence, that’s not our only reason for being here. We each, to varying degrees and extents, have our own motivations for being here. In my case, I’m here to learn, and I’m here to interact with people who share my love for AI (in ALL of it’s forms, not just the “latest and greatest”), and to share my views and ideas, hopes and dreams for the field in general. Of course, I can’t speak for everyone here, but I can certainly speak for myself. And I feel that I can say with absolute certainty that you, Sir, are NOT in the wrong place. Your views and opinions are every bit as valid and worthwhile as mine, or anyone else here, for that matter. And just because we don’t always agree on things, that doesn’t mean that we can’t be friends. It’d be a pretty boring place if everyone agreed with each other 100% of the time, don’t you think?

I should also point out that Merlin had a good point earlier when he advised us not to confuse the AIML standard with AIML interpreters, or AIML content. What I was referring to, when mentioning “maturity”, was the AIML standard itself, which has been pretty much unchanged for ten years now. I’m well aware of the fact that developers, like Liz Perreau (who created Program O and is working even now to release an improved version), and botmasters like yourself and Steve, and even myself, are constantly working to make improvements in our respective areas of expertise, and will continue to do so for the foreseeable future.

I think that, for the most part, the use of AIML isn’t as wide-spread here as with other communities. Most of us here have, at one point or another, used AIML, so, generally speaking, we’re pretty familiar with it. Most of us, however, no longer use it as a basis for our chatbots. Does that constitute a stigma? Possibly, but I don’t think so. Certainly there are some who may want to compare AIML to a “box of old, worn out crayons”, when compared to the “masterpiece” that is their favorite “pet project”, and they’re entitled to their opinion. But should that color one’s perceptions and impressions of the attitudes of the entire community? Not for a single second.

Thunder Walk - Jun 29, 2011:

Not to argue, but for the sake of discussion…

I think that the word “better” is a relative term in the view of lawyers, scientists, or technocrats of any kind.  Down here on the ground, we know what better means.  When most people say, “better,” they mean, “better for me”.  Ambiguity is in the eye of the beholder.

When the comment was made, “...there is a bit of a stigma about using AIML,” the response was, “...AIML is pretty much a “mature” technology, with little in the way of improvement or innovation over the past several years” and “When a craftsman becomes comfortable with his tools he relies on them and is hesitant to change. I would doubt that even if there was something better out there that current botmasters would move to it…”

I’m confident that someone, somewhere, is working on everything to improve, expand, and modernize it.

This may be the crux of our difference in viewpoint. I am not as confident that AIML will evolve with the times. It works, and there does not seem to be much interest from the AIML standard bearers to update the language. The last spec is dated 2001 and is listed as a “working draft”. http://www.alicebot.org/TR/2001/WD-aiml/

Others using AIML have suggested that it could be improved.
http://www.chatbots.org/ai_zone/viewthread/477/
http://www.chatbots.org/ai_zone/viewthread/473/

Richard Wallace - Apr 15, 2011:

Identifying a keyword in anywhere in an input sentence typically requires four AIML categories: the cases of the word being at the beginning, in the middle and at the end of the sentence, and the case of the word by itself.

The keyword and wild card problems are often referenced as things AIML users would love to have changed.

Thunder Walk - Jun 29, 2011:

My point is, until something better comes along, we’re all relegated to employing something less… what’s available.  Still, I understand that pushing the envelope is always interesting and probably more fun.

I disagree that people continue to use AIML simply because they’re lazy or comfortable.

A craftsman does not use his tools because he is lazy or comfortable. He uses what works. But, his investment in learning and becoming skilled with his tools makes him less likely to change to something new since he would have to start all over.

Thunder Walk - Jun 29, 2011:

Not everyone is, or is interested in becoming… a computer scientist.  But, I’m also confident that when something easier, faster, smaller with more content comes along, current users will migrate to it… when it’s better for them.  And, therein lies the key.

In building technology products, you can often find marketing guys talking about the adoption model. Some people are willing to try new technology before it has been fully proven as “better”. These are called “early adopters”. “Laggards” are at the other end of the spectrum and must be convinced that there are other users that have proven the utility and every feature of the product is better (and even then a percentage will not adopt the product).

The range of users across the spectrum will adopt when the new product crosses their “benefit” threshold. The question is always how to prove another solution is better.

As a thought experiment: Imagine for a moment the there is an new language next year that is “better” than AIML by any definition that you want to set. The only limitation is that there is no “content” available in this new format. Would it be possible to prove to you it is better?

Thunder Walk - Jun 29, 2011:

However, if only futuristic thinking, and a devotion to whatever is next is the soul purpose of this forum, and only commentary regarding yet another… and another… and another Proof of Concept bot is welcome (or understood) then, Steve was correct.  The stigma regarding AIML seems to exist, and I’m in the wrong place.

This forum should be a place for both “pathfinders” and “colonists”. I don’t think the “AIML stigma” is about AIML as a tool. Obviously it is a useful tool. Historically it may have been the most productive tool. We can argue about whether or not it is the best tool currently available.

I think the “stigma” is that the language is “stagnant”. Thunder, as an AIML advocate, if you had control of the AIML language spec, what are the top 3 things you would change/add?

Merlin - Jun 29, 2011:

... as an AIML advocate, if you had control of the AIML language spec, what are the top 3 things you would change/add?

I can think of two at present, any others, I’m confident, will come to me in the middle of the night, or at some inopportune moment.

The first has to do with words we’ve come to label as being “hard-wired,” or “hard-coded” at the server, or perhaps maintained in a substitution file that’s inaccessible to the botmaster.

The letter “U,” is solidly associated with the word “YOU.”  The word “REALY” a misspelling of “REALLY” is converted as though it were correct, and the word “WAHT” is hard coded in all it’s various forms (WAHT, WAHT *, WAHT IS *) to match “WHAT”, thereby accepting misspellings, or typos, as being correct.  The typos “HELLP” and “HELLP *” are hard coded to reduce to “HELP”.  The result is, if someone meant to type “HELLO CHATBOT,” but instead, accidentally typed “HELLP CHATBOT,” the reply is, “Can you ask for help in the form of a question?”  There are probably other examples, but I’m unable to recall all of them.

For various reasons, I’d like to have control over all words, but it’s forced some of us to find creative solutions to overcome those issues, when possible.

The second has to do with with the wildcard “_” (underscore) superseding everything including an exact match.  If you have 2 categories,” _ ROBOT” and “ARE YOU A ROBOT,” the first one will always take precedence even if the user says, “ARE YOU A ROBOT”.

But, back to the original topic… Is there anyone reading this who denys that there’s little interest or faith in AIML in this forum, evidenced by the lack of commentary when the subject is raised, or that virtually no one brings it up in topic… yet, there are dozens of postings followed by an equal number of replies concerning a one of a kind bot that does little and is utilized by no one but the person who created it?

I don’t pretend that AIML does more than basic pattern matching, but with the use of tags such as < person> and <srai> you can accomplish some clever responses.  True AI won’t be accomplished or made available to the public anytime soon.  If that’s going on anywhere, it’s in the battery of computers several levels below ground in the Department of Defense buildings.  I think the notion that we’re breaking new ground here, or that somehow if we combine the collective knowledge and brain-storm that something magic will happen, reflects a kind of snobbish elitism that afflicts great minds in high places.

While AIML isn’t perfect, I’ve yet t see anything better, and I doubt we will anytime soon at this level.

Thunder Walk - Jun 29, 2011:

Is there anyone reading this who denys that there’s little interest or faith in AIML in this forum, evidenced by the lack of commentary when the subject is raised, or that virtually no one brings it up in topic… yet, there are dozens of postings followed by an equal number of replies concerning a one of a kind bot that does little and is utilized by no one but the person who created it?

Actually, I thought this board had a pretty good following from AIML people. It attracted the best AIML guys I know (you, Steve, Dave, Gary, Dr Wallace, etc.) and has also bought in people and projects I never heard of:

http://www.chatbots.org/ai_zone/viewforum/2/

Leo Ktaba’s enhancement suggestions for AIML are interesting.
David Newyear’s catbot is award winning.
There are a number of AIML based thesis projects that come up on the board all the time.

Thunder Walk - Jun 29, 2011:

True AI won’t be accomplished or made available to the public anytime soon.  If that’s going on anywhere, it’s in the battery of computers several levels below ground in the Department of Defense buildings.  I think the notion that we’re breaking new ground here, or that somehow if we combine the collective knowledge and brain-storm that something magic will happen, reflects a kind of snobbish elitism that afflicts great minds in high places.

I challenge that assumption and raise you Apple, Adobe, Microsoft, Google, and Facebook. Many a great new technology has been born of “4 guys a dog and a six pack”.

Thunder Walk - Jun 29, 2011:

While AIML isn’t perfect, I’ve yet t see anything better, and I doubt we will anytime soon at this level.

Now that is a challenge I may just have to take you up on.

When you think of best, you think of competition and the winners of contests.  AIML entries typically show well in the current gamut of challenges, but they don’t win as often anymore.  Definitely there are better bot authoring tools now.

AIML is frozen, not because the people who employ it haven’t had suggestions for changes, but because of its history with adopting changes into a standard (something hinted to by the last attempt still being a draft.)

The problem continues because these kind of tools don’t address the fundamental issues in creating a bot.  The idea of creating a character or personality, that is, specifying kinds of behavior, without a situation, a plot, a narrative if you will, turns into a massive effort. Especially when it is taken a volley at a time, a play with lines cobbled together from many other scripts.

Until a bot builder can give just one, or only a few analogies, of distinguishing scenarios exemplifying each major bot characteristic and the (future’s) tool then extrapolating those summaries into how the bot interacts with the world and grows in character, we’ll still be left with a most difficult task of making automatic shallow responses to generically matched pattern of inputs in order to create a suspension of disbelief (the impression that the chat bot is not a stupid program.)

It is as simple as asking how you would describe your friend.  Would you say using AIML is the best way?  Or would you rather just circulate some “gossip” to the authoring mechanism?

BTW, being able to formulate a whole personality from not too many illustrations would not be a stupid program!

Hello I am new member in this forum. My language is Spanish, and I have not much control with the English language, I am using a translator to communicate with you. I apologize if I not understand well.

My first question relates to this thread. I am creating my first chatbot in pandorabots. And I want to change my html page, but I have no knowledge of html.

I tried to copy the source code of the pages that appear in the links, but do not serve me, because my chatbot fails.

Someone can help me create my own html page for my chatbots?.

Hello, Aldrin, and welcome to chatbots.org!

I can help you create an HTML page for your chatbot. Please send me an email using the link below my avatar image, and we can talk about your needs.

The conclusion that I have came to in regards to AIML is that it reminds me a lot of the early days of HtML. A utility mark up language at best. Yes, you can impress with lots of markup and more mark up, but all your really doing is building an interactive database facts and not true AI. Having the answers to many questions does not prove innovation just a dedication to hours of input to impress. HTML has evolved to the point of replacing other technologies like Flash and doing so more efficiently. Possibly this could be the case with AIML, but personally I don’t think so.

Neural Networks shows much more promise in the right direction, but still there is more R&D needed. At this juncture, useful applications with piezo AI are needed to pave that pathway for a more broader acceptance of AI. Artifical Intelligence is the next big thing in terms of user interaction with devices and applications. We are just now starting to explore the possibilities, and it is exciting.

With all due repect Laura, anyone who describes AIML as a “utility markup language at best” and “all you’re really doing is building an interactive database of facts” has obviously never tried it beyond the basics. This would be like saying a concert piano is a child’s toy because you are only capable of playing a simple tune with one finger.

While I have to agree with Laura’s assertion that AIML is not “true” AI, I also must side with Steve in that AIML is far more than just a “utility markup language”, and that it also extends well beyond the simplistic notion of an interactive database of facts. I may be a bit biased, but I see great potential in AIML that only a very few people are taking advantage of (Steve being first among these folks). Once my schedule loosens up a bit this spring, I’ll see if I can’t provide some proof of this, in the form of a new chatbot project, named Morgaine. She’ll have all new, personally coded AIML responses, made just for her, and will run on Program O’s version 2.0 platform, and I’m hopeful that she’ll far outpace Morti, and maybe even give Mitsuku a run for her money.

Clearly there’s no actual “thinking” going on with AIML, and yes, it’s very much like HTML, however that’s one of it’s advantages.  If you can do one, you probably can easily handle the other.

The payoff of AIML is that there’s a lot of bang for the buck.  Obviously, it’s free, but more than that, it’s economical in that there’s not a lot of work involved, and it’s not necessary to have voluminous files to create something that appears functional.  Anyone can learn it without first acquiring advanced degrees in mathematics or computer science.

But, any botmaster babbling in AIML can tell you that by combining patterns, and the strategic placement of a few tags, your simple AIML bot will sometimes offer up surprising replies that you know you didn’t put there.

As for “Having the answers to many questions does not prove innovation just a dedication to hours of input to impress,” I don’t believe I’ve ever read where anyone claimed that it’s something that it’s not.  True, AIML is comprised of questions and answers, but show me a system that isn’t.  Even Watson had to be taught answers to questions for it’s appearance on Jeopardy, and it didn’t do all that well during the first videotapings of the TV show.

And, just like more complicated systems, AIML has the ability to pick up on keywords and phraises (just like Watson) and can then search other data bases such as Google for a reply… an image… or an audio file.

Okay, I knew with my statements about AIMI I would stir the pot a bit. I did not mean to come off so trival about AIML other than my opinion that AIML is never going to be the dominate technology when it comes to the future of AI. Comparing it to HTML in its structure as a markup language is a fair assessment since it is tag based. Someone like myself that had worked with the earliest versions of HTML, it’s natural to make the comparison between the two. I will agree that AIML can do much more than format and display answers, but still has its own limitations as well.

With the above said, I still defend the fact that AIML is a transit language and in looking back a few years from now, it will be looked upon as a basic attempt to mimic true AI. It is however, an easy to use basic tool for building Chatbots for those of wish to dabble with AI and build entertainment applications. You have to admit that researchers working in the field of AI are not pounding AIML in their laboratories.

No worries, Laura. We each just wanted to try to correct some minor(?) misconceptions about AIML; more for future readers than to pick on you. Your assessments about AIML being a “transit(ional?)” language, and that it’s decidedly more geared toward entertainment than research are fairly spot on. But as you alluded to, it’s an effective tool for what it’s intended for.

I know you guys were not picking on me.

I have no direct experience with AIML other than a peek at the structure which gave me a fair insight into both it’s merits and shortcomings. I have to remember that Chatbots are the focus here and not AI applications in general. My head has been stuck in the clouds (literally) as I am developing a data scraping engine which is key to the Slide Agent project.

Playing around with AIML sounds like a lot of fun, so I know why so many are attracted to Pandorabot platform.

hours of input to impress

Isn’t that true of the human brain as well?  Even people have to learn, memorize and sort through random facts.