Do computers think??? Does aiml Allow computers to think???

Postby joehop67 » Sun Jul 01, 2012 5:31 pm
Hello I am a new user. I have been developing with AIML for a little over a year now. I was just doing some research and i would like your opinion on something. Do we remember our old friend alan turing (who we can undoubtedly say may be the father of common ai) he said that if a computer can hold a conversation with a person without the person seeing the computer any different from a real human then (in any reasonable sense of the word) the computer is thinking. Well now thinking into AIML the computer has to match text input (or voice input) to patterns in a code set right??? well if it has to match that then it is in fact thinking. Think about how we think: as i am typing this i have to match the letters, words, keys, etc. to things matched into my memories right??? well how is that any different from the ALICE bot checking its “memory”(AIML) for the correct response. It is what we do everyday we have to match sentences people say all the time to make sure we get it right. Its no different than if we where chatbots. In our modern world we really just very advanced chat bots….. This is all opinion based and i would really love to hear what you guys think.

~Joseph

Hello, Joseph, and welcome to chatbots.org!

Wow, this is one doozey of a case of Deja Vu! It seems that I just read this on the ALICE AIML forums. That’s ok, posting in both places is a good way to ensure better coverage of your thoughts and ideas. I say good on ya!

Anyway, to also repeat my post form the afore mentioned forum, I offer this:

Respectfully, I have to disagree at this point in an absolute sense, mainly because of the nature of how AIML works. For the most part, if you have multiple conversations with a typical AIML chatbot with a given user typing in the same inputs (bear with me here, there’s a point), you have a better than 90% chance of getting the same responses. This is because of the highly “scripted” nature of most AIML sets in existence today. Sure, there are some AIML chatbots out there that this scenario doesn’t strictly apply to (Mitsuku comes to mind, among others), but by in large this bears out. Even my chatbot, Morti is pretty limited in this respect, though I like to think that he’s slightly more “random” than most other bots. The reason for this is that the overwhelming majority of botmasters fail to take full advantage of the full potential that AIML offers. This is an issue that I’m hoping to fully address with a new chatbot that I’m working on, named Morgaine. My goal for Morgaine is to fully utilize “personality gauging”, by categorizing a user’s mood through the use of “keyword analysis” (there’s nothing fancy here; it’s just that certain keywords will trigger the alteration of certain variables that will be used to “guide” Morgaine’s responses), and setting certain variables to monitor a user’s “state of mind”, and (more importantly) to vary her responses based on this data. It’s a huge undertaking, and will end up with Morgaine’s AIML database containing several hundred thousand categories, with a large portion of that being special categories that will process the various “state of mind” settings, factor in her own “mood”, and formulate her responses based on the current “emotional environment”. Even then, I still don’t think that Morgaine will be “thinking”, though she’ll certainly seem to be (I hope!).

Of course, this is just my considered opinion, and as such, is no more (or less) valid than your own, so feel free to ignore me.

Thank you for the reply… LOL i found this place after realizing that not many people use the ALICE/AIML forums anymore. I am in 100% total agreement with your opinon(which i know makes like absolutly no sense because of what my original post said). Your chat bot sounds very very awesome. I would love to help and i could but i am afraid that i dont think i know that much about AIML. Of course i know the basics but i would have no clue what to do if i was helping you. I wish i could help you (i could (of course) help with the basics and probably reduce the time amount of your project. but i would never be able to help you with the more complex actions).

Thanks a lot for the reply.

~Joseph Hopkins

Well, right now, Morgaine is little more than a technical outline and a (very small) collection of test categories, that are a “proof of concept” for some of the advanced ideas that I have. I want to try my best to make her a “purely AIML” chatbot, but some of the more complex stuff may require me to “cheat” at some point. I’ll have to see.

What I mean by “cheating” in this context is writing certain addon scripts for the chatbot engine that I use, called Program O. I’m a part of the development team for version 2.0, which is currently in public beta testing, and I’ve helped to make it fairly simple and straight-forward to create addon scripts for it. I think I can use just AIML to index certain keywords for emotional states, such as self confidence, happiness, and others, and to modify certain variables that will help guide the responses, but this is still pretty much in the research and experimentation stage, and so isn’t very reliable, and may end up requiring the use of database manipulation to get it to work. I just don’t know yet.

My biggest challenge with this problem won’t actually be the technical aspects, but the vast amount of “ordinary” categories that seem to push my creative writing skills beyond their limits. After all, one of my primary goals for Morgaine is that there won’t be a single line of AIML that comes from any other AIML sets, so she’ll be 100% original, and will (hopefully) never be accused of being an ALICE clone. So if you’re willing to help, I’ll certainly be willing to accept it.

One of the things that I plan to do when I actually get down to writing out her responses is to take huge quantities of notes about my conversations with other folks; not so much for what is said, but instead, to note certain emotional states, and how these states may have influenced the conversations, and which emotional states may have been in turn influenced by the conversation, as well. It’s these types of interactions that I’ll need to study, so that I can possibly “capture” them in some sort of programmatic way. I also plan, at some (soon?) point to compile a list of “emotional keywords”, to analyze the effect that their use in conversations has, both on the conversation itself, and on the people participating. This is probably somewhat above my “educational pay grade”, but I think I’ll enjoy the challenge. I never took any sort of psychology courses, but I’m not going to let a little think like complete ignorance get in my way.

I could probably go on talking about this project for days, but I think I’ll shut up for now, since it’s a bit off topic.

Okay.. well if i can help by just having some simple aiml and others (i may be able to find some “emotional keywords”) then i will be more than happy. I would love to hear more about your project so email me at: annanamousejoe(at)live(dot)ca with more details. And your working on program-O ? thats awesome !!!! hey can you give me any updates on the addition of the CallMom aiml set??? im trying to make a DLA for your android phone.

So yeah i would love to hear more about your project and what i can do to help.

Thanks

~Joseph Hopkins

{email address altered to help prevent email spam}

Added support for the CallMom tags should be implemented once we can get the script’s current bugs worked out, and move it out of beta and into production. Hopefully in the next few weeks. Lately it seems that fixing one bug produces another, so sometimes it feels like an uphill battle, but we’re making progress. There are actually only two tags that need to be added to Program O (and one isn’t really a CallMom tag at all), so it shouldn’t be too difficult to implement. We just want to get the script all neat and pretty first.

thats good to hear but how do are you guys gunna make it so that.the oob tags actually control the phone ???

See, the nice thing about that is that the CallMom app is what parses the XML output that goes between the tags, so that all Program O has to do is treat it as text, and output it straight to the phone. Thus, there’s really little for Program O to do, except for removing the <oob> tag itself. If there are any AIML tags within the tags, they get processed as usual, but the rest is just text to Program O. The other tag, <explode>, will probably be more of a challenge, but not significantly so.

Oh i see. Im glad to hear that its not going to be that much of a challange. I cannot wait till the next release !!! im soo exited. Now if only i could figure out some way to make program-O work within an app. The most troubling part is the speech input/output(now THATS going to be a real challange) Maybe(if i figure it out) i could turn your new chatbot into a mobile app(if thats something you would be interested in) but the problem is… I have NOOOO clue how to program in java…. LOL well i guess i better get studying -eh lol

Making Program O work within an application is as simple as can be. Version 2 has a built-in XML interface that returns a “mini-XML document” that the app can parse into a text response that the app user can read. Depending on the language/platform of the app (and the difficulty of the questions), I may be able to assist with answers to specific questions, but bear in mind that my IOS skills (iPhone) are virtually non-existent, and my Android skills aren’t “top of the line”, either.

Well even your non-“top of the line” Android skills Are still like 50 times better than mine lol. I have made an android app before but that was using an online servus. And thank you for editing my post :D I can of course run it in call mom (if i run it in pandorabots) but then theres no way to monotize it. LOL my actual insperation was HAL 9000 from 2001: a space odyssey. The voice would be simple to make but the app its way to hard. Some people seem interested but not the “right people” you know what i mean??? I thought it would be simple to find an android developer and it is but they want cash of course and the best i can promise is 50% of the profits (Which of course no one will take cause the app could be a flop) And by the time i learn to make the app android may be replaced by something better. But what ev’s i guess i have to start somewhere.

Never let the possibility of failure stand in the way of what you want to do. It’s not “failure” till you give up. Also, never pass up an opportunity to learn something new, especially if it helps you toward any given goal. Learning is living, and without either, what have we got?

And by the way, if you’ve only made a single Android app, that’s more than I have. I’ve only installed the Android OS on my desktop machine for testing purposes, and read a couple of online manuals. However, I’m a VERY quick learner, and Google is my friend, so I don’t see that as a problem.

LOL it just takes a long time to learn that kind of stuff and i just dont want all that research to be for nothing. I know you can use C# to make android apps but im pretty sure you have to download another thing to make that app work. Plus im not all that expiranced with c#. IDK ill probably start learning java but ill still be looking for a developer to make it. Im just not sure what to call the bot. I already have a bot that runs in CallMom but i want it to be like HAL cause hes like my favorite movie character ever. What do you think, should i make a new bot ??? or just leave it as is ???

I say that if you want HAL, then don’t stop till you have HAL. The biggest reick to making your own chatbot with unique responses and a personality to match is in the creative writing aspect. In essence, when you create a chatbot, you’re writing a story that doesn’t “unfold” in the same way that a book does but still has to be entertaining, engaging, and true to it’s purpose. That’s no easy task, but in the long run, the results are well worth the effort. Since the character/persona of HAL is actually a rather simple one (How many lines did HAL actually have in the movie, after all?), it would be rather easy to create, relatively speaking. You can take the base ALICE set and use the reduction files, and most of the files that make up the “structure” of a chatbot, but that don’t contribute to the bot’s personality as a start. That’s nearly 2/3 of the AIML files, size-wise. Then take the remaining files, one at a time, starting with the smallest (starting small is far better than the other way, trust me) and working your way through the files. Just start with the first category, look at the response, and ask yourself, “what would HAL say here?”. If a category doesn’t seem to fit at all, don’t be afraid to delete it and move on. Once you get into a rhythm, it will go by fairly quickly. If you need help or advice with this, you have my contact details, so feel free to holler, and I’ll help where I can.

By the way, folks, I know that this thread has become a bit of a “garden path”, but I don’t mind, so there.

Garden Path??? and I know HAL didn’t have a lot of lines but look at his (presumed) IQ hes very smart controls everything on that ship. The HAL Bot would be fairly easy to make but the full extent of HAL is out standing. There was a guy who actually did something like that look up the JARVIS PROJECT its really interesting. the app i wanted to make was basically the HAL for your phone. (hey you wouldnt happen to know where i could find how to use a TTS software with pandorabots ??? I had an idea of how i could make a “HAL” for the computer. Have you ever tryed using text to speech software with an AIML bot??? HAL would be easy all you really need is an image of the hal interface and someway to input it then you would have to place in the tts software onto the webpage and we are home free. The only problem is i have never used tts software in my life lol.