Hans—your point is well taken regarding my seemingly inconsistent message.

When I started my project, in late November 2008, I really did know exactly what I wanted :

a better chatbot.

A bot that:

1. can understand your input—and not in a shallow, template based way.  Not an Eliza “pick keywords” way.
2. derive a meaning, that is, what you meant
3. know how your last input relates to the conversation as a whole.
4. store information, and retrieve it via NLP.
5. act as a tutor for students learning things like electronics
6. learn new facts and relationships
7. learn by example, procedures.
8. discuss and troubleshoot, perhaps to be used in a call center.
9. just for fun .. chitchat.. but yet really know what you are talking about.

and if stuff like that is achieved, then I may try more abstract things.  If it does not succeed, I will investigate other approaches.  There is no way we can no for certain that any one of us will be successful.  Are you 100% sure you will be successful ?

  So maybe some of my latest ideas are starting to approach this whole ‘strong AI’ perhaps, I don’t know.  But I don’t really want to be associated with that term.

My bot won’t be able to do things that most would really consider true AI—it won’t do things like invent calculus or write the Grand Unified Theory—yours will, I can’t wait to see it!

What kind of a list of things do you want your bot to do ?

Well, to be honest, I’ve already forgotten exactly how I voted, but that’s more of a “CRS” problem than anything else. My personal perspective, however, varies from my “professional” perspective, due to the very nature of the project that I’m involved in, which is basically trying to take a simple AIML bot (pattern based input/response correlation), and making it “evolve” into something more. And, like Hans, I’m still in the “discovery phase”, where I’m trying to learn all of the new concepts and avenues of research, sorting through and picking the ones I think will best suit my goals, and then contemplate how best to integrate them.  Thus, while I’ve made at least some improvements to Morti’s script (extended spell checking/substitution, for example), he’s still only performing (I forgot the term Gary used. DRAT!) ... “data manipulation” (for lack of a better term). This, of course, will change over time; perhaps over a long time.

Hans Peter Willems - Feb 15, 2011:

I did not register to go around and just post how much in awe I am from everyone’s accomplishments (although I seriously respect everyone’s efforts here).

I’m not entirely certain if this is directed towards anyone in particular, but I get the faint impression that it may be directed at me. Whether or not it is, I thought that I would point out that, given my “amateur status” here, I find a lot to be awed about, and that’s all to the good. I wouldn’t like to think that anyone felt that I was heaping on praise because I was “sucking up”, or anything like that. Frankly, I have no use for people who behave in that fashion. But I do like to let people know when I think they’re doing a good job, and I do seem to be a bit effusive with my praise. That’s ok, though. Positive reinforcement and psychological incentive are both valid ways of boosting efficiency.

Hans Peter Willems - Feb 15, 2011:

Btw, I would really like to hear from the other people here that voted against ‘grammar first’

I was one of the ones who said I don’t care about grammar. This is specific to my development with Skynet-AI (not other things I might build with JAIL).

I work with concepts. Most concepts are built into a neuron. Some neurons can be bundled together and are triggered by other neurons. Neurons are normally fuzzy but can be crisp.

Skynet-AI takes any input and tries to provide an appropriate response. Depending on how well (or not well) it understands the input, it tries to forward the conversation in an intelligent manner.

Concepts include, math, questions, small talk, yes, no, hello, etc.

Victor Shulist - Feb 15, 2011:

Are you 100% sure you will be successful ?

That of course depends on how you define ‘100% success’. But I’ve never been accused of ‘not being ambitious’

Victor Shulist - Feb 15, 2011:

yours will, I can’t wait to see it!

Victor, please refrain from derogatory statements like this one. It’s things like this that turn civil discussions into evil mayhem.

Victor Shulist - Feb 15, 2011:

What kind of a list of things do you want your bot to do ?

I started out about two years ago with building a chatbot in AIML. I was using Program-E which had several serious bugs. From these experiments I moved to ‘strong-AI’, simply because I became bored with the AIML-based results. It made me long for something more ‘involved’, so I made the jump into research for a strong-AI mind-model.

So what do I expect from my model (or: what defines success for me):

1. Having a mind-model that is capable of storing information that is related to ‘tacit knowledge’ in humans.
2. Having a software implementation that is able to ‘learn’ concepts and learn itself from those concepts, and so forth, based on the above mentioned mind-model.
3. Have the AI-mind come up with at least one concept that can not be ‘directly’ related to information that was input by the operator, i.e. showing reasoning based on comprehension (the ‘spark of consciousness’?).
4. Being able to have a conversation as you would have with another human, i.e. being able to pass the Turing-test.

Dave Morton - Feb 15, 2011:

I’m not entirely certain if this is directed towards anyone in particular, but I get the faint impression that it may be directed at me.

It was not directed at you Dave, it was meant as a very general statement. Although I see nothing wrong in showing some admiration where appropriate (I’ve done that myself), nothing is gained from continually ‘stroking’ each other’s egos.

Innovation comes from conflict

Thus, the reason why the vast majority of overall technological progress happens during times of war. However, I’m of the opinion that there needs to be some positive reinforcement involved to keep things balanced. But point taken, so I’ll try to moderate my praise a bit.

Hans Peter Willems - Feb 15, 2011:

Innovation comes from conflict

That’s a bit of an over-simplification.

But conflict is a good thing in some cases.  For example, having your ideas attacked (I don’t mean to be defensive here, really), and then asking yourself if you think your current path is the right one, and getting a definite YES in return, well, that’s a good thing

Grammar still appears to be in the lead.

Dave Morton - Feb 15, 2011:

We seem to be heading back toward the whole “Gary vs. Victor” scenario again, and I think that’s counter-productive. {sigh!} It’s my considered opinion that Victor’s work on a grammar engine is a valid avenue of research, though hardly complete, that will achieve some level of success at some point as part of a larger package.)

That’s a good analysis Dave. 

You see, the main philosophy I am following in my development of CLUES is as follows.

First off, it is an incremental approach.  Goal based. 

So I don’t start off with some nebulous idea and use words like ‘concepts’, I identify a very specific and clearly defined objective, then I work my way up.

In order to talk with a bot, it has to understand you.  You generate parse trees.  From parse trees you generate semantic trees.    Semantic information is added to the tree in order to know which (of the many) trees is the one that makes sense.

That is first step, and has been accomplished.

Now, of course there is much more to do like you say, and yes, absolutely the parser engine is only a part of the larger whole.

I have much more to think about, much more functionalities that need to be added (spell check for example).  There will be many algorithms in addition to parse tree generation and disambiguation.

The system can be given a 30 word complex sentence generating 10,000 parse trees, use semantic inference to decide which one makes sense, (in 1 or 2 seconds)  and answer all kinds of questions about it…. no, that is not the “end all”. . in fact , only the beginning, but it sure is showing promise.

But Hans, you yourself stated you are in the same situation (that you are not complete your theories and have much to work on), in fact you have written no code at all and thus have no example I/O at all… yet I am on the wrong track ??!!

Dave—cancel my account.. I have work to do

Victor Shulist - Feb 16, 2011:

Hans Peter Willems - Feb 15, 2011:

Innovation comes from conflict

That’s a bit of an over-simplification.

But conflict is a good thing in some cases.  For example, having your ideas attacked (I don’t mean to be defensive here, really), and then asking yourself if you think your current path is the right one, and getting a definite YES in return, well, that’s a good thing

Grammar still appears to be in the lead.

By conflict I don’t mean ‘fighting’.

By conflict I meant ‘colliding views and ideas’. When you are reasoning within your own mind, there is internal conflict as well

Victor Shulist - Feb 16, 2011:

So I don’t start off with some nebulous idea and use words like ‘concepts’, I identify a very specific and clearly defined objective, then I work my way up.

So you actually say that only the engineering approach is valid, and working on research is ‘nebulous’ ? Maybe you should tell the academic world that they better stop formulating theses and researching to prove them, as that is all ‘nebulous’.

Victor Shulist - Feb 16, 2011:

In order to talk with a bot, it has to understand you.  You generate parse trees.  From parse trees you generate semantic trees.    Semantic information is added to the tree in order to know which (of the many) trees is the one that makes sense.

So you say that this is the only way to go, right… this is ‘how one goes about it’ ?

Victor Shulist - Feb 16, 2011:

But Hans, you yourself stated you are in the same situation (that you are not complete your theories and have much to work on), in fact you have written no code at all and thus have no example I/O at all… yet I am on the wrong track ??!!

First, not having code and not having example I/O does not invalidate my approach, only results from research can do that.

Second, I never said you are on the wrong track; I discuss with you that I think something is missing in your model to accomplish what you say you want to do. Maybe just a week ago you said you where only interested in building a weak-AI chatbot. I said then that your work on that was pretty impressive. But then you moved onto ‘higher reasoning’ and ‘maybe strong AI’ and started discussing how that could emerge from your current approach. That is what I’m contesting, and in doing so we have a heated discussion about our viewpoints. I see nothing wrong in that. Maybe you are right, maybe I am, but in the meantime why not learn from each other’s opposing views and ideas.

I post my ideas here to solicit responses; I learn from those responses, it’s as simple as that. The discussions so far has helped me shape several parts of my thesis already. As I said, I don’t want praise, I want my ideas to be contested

So you actually say that only the engineering approach is valid, and working on research is ‘nebulous’ ?

Maybe what Victor is trying to say is that, for as long as you are trying to bake air, your are only going to be eating air. Right now, you are claiming your air to be better then our cakes! Personally, I prefer something substantial in my tummy.
An idea by itself is worthless, make your idea work, then things become interesting!

Jan Bogaerts - Feb 16, 2011:

Maybe what Victor is trying to say is that, for as long as you are trying to bake air, your are only going to be eating air. Right now, you are claiming your air to be better then our cakes! Personally, I prefer something substantial in my tummy.
An idea by itself is worthless, make your idea work, then things become interesting!

Hear hear!

Implementation is not only the proof in the pudding (er, cake) but it also helps clarify ideas and theory.

I resent the idea that, because I’m formulating a thesis without (at this point) coding stuff into ‘something that does something’, means I’m trying to ‘bake air’.

I must say that, with all the bright minds here on the board, I didn’t see that one coming.

Do you perhaps have a paper or an essay? or a mathematical proof (though someone else would have to read that)?

Jan Bogaerts - Feb 16, 2011:

Do you perhaps have a paper or an essay? or a mathematical proof (though someone else would have to read that)?

Not yet, because THAT is what I’m currently working towards (I stated this several times already). But, although I’m working from my own premise, I already found several papers from respected researchers, and other sources, that are pointing to the validity of my ideas. The NELL project for example has many ideas that are very close to my own model. It is not EXACTLY the same, but close enough to lend at least some validity to my own research so far.

About NELL: http://www.nytimes.com/2010/10/05/science/05compute.html?_r=1

The main difference between those projects and my own model is that I want the model to be able to describe ‘tacit knowledge’. Describing such information in such a way that it can be processed by algorithms is a big hurdle to take. After that, writing those algorithms is, in comparison, a small task. Hence my current focus on developing a model that can describe those things, instead of coding ‘stuff’.

And I like to point out again; I work professionally in the field of applied Knowledge Management systems, so the focus on the describing model is very logical for me. Things like building taxonomies and ‘cognitive navigation of information spaces’ (I’m working with subsidized funding on development in that area) are daily tasks for me.