•  * * * NOTICE : This is a philosophy free area

Since nobody really knows what AI should be (not even what intelligence really means), I suggest just focusing on a chat bot that can actually perform useful tasks and, through interactive conversation learn new things, and be basically a Google killer (unless Google buys it )

This is SNAI Research.

Let’s make a wish list here… give me your sample “dream conversations” you would like to be able to have with your bot, and we’ll make it happen.

I think something like the following sample conversation would be a very good start.

Person: Pick up a big red block.
Computer: OK.
Person: Grasp the pyramid.
Computer: I don’t understand which pyramid you mean.
Person (changing their mind): Find a block which is taller than the one you are holding and put it into the box.
Computer: By “it”, I assume you mean the block which is taller than the one I am holding.
Computer: OK.
Person: What does the box contain?
Computer: The blue pyramid and the blue block.
Person: What is the pyramid supported by?
Computer: The box.
Person: How many blocks are not in the box?
Computer: Four of them.
Person: Is at least one of them narrower than the one which I told you to pick up?
Computer: Yes, the red cube.

Ahhh Yes…. Block World   The classic ‘70s program.  It *did* show a computer program could converse in natural language effectively, didn’t it?

Victor Shulist - Sep 15, 2010:

Ahhh Yes…. Block World   The classic ‘70s program.  It *did* show a computer program could converse in natural language effectively, didn’t it?

I was of course quoting from a dialog with SHRDLU which was written by Terry Winograd around 1970. http://en.wikipedia.org/wiki/SHRDLU

Unfortunately it failed to show that a computer program could converse in natural language effectively because it was restricted to such a narrow toy domain, the block world. The entire program was “hard-coded” and lacking any kind of abstraction mechanism for language and grammar, it could not be extended to other areas. It was a dead end and serves as an excellent demonstration of what not to do.

Here’s another sample dialog for consideration. In this conversation, “Doctor” is a real computer program which was developed using a modern general purpose natural language processing system. In addition to providing medical advice it is being deployed in a number of diverse real world applications. Although I’ve disguised the name of the program, it shouldn’t be hard for anyone who is used to doing a little research to track the publications that describe it using Google.

Patient: Can I take an aspirin?
Doctor: No, you are taking Celebrex and Celebrex interacts with aspirin.
Patient: Oh. Could Celebrex cause headaches?
Doctor: No, headaches are not an expected side-effect of Celebrex.
Doctor: Do you have a headache?
Patient: Yes.
Doctor: How severe is your headache on a scale of one to five.
Patient: One.
Doctor: Your doctor says you can take Tylenol for mild headaches.
Patient: Ok, I’ll try that. Why am I taking Celebrex?
Doctor: You are taking Celebrex to help with your arthritis.
Patient: Should I take one now?
Doctor: Yes, you are taking Celebrex every morning.
Patient: Is there anything else I need to take now.
Doctor: Yes, you should take your Prinivil.
Patient: Which one is that?
Doctor: It’s the yellow one. [shows picture]
Patient: Could that be causing my headaches?
Doctor: Yes, headaches are an expected side-effect of Prinivil.
Doctor: Tylenol should help.
Patient: Ok, thanks.
Doctor: You’re very welcome.

Andrew Smith - Sep 15, 2010:

Person: Pick up a big red block.
Computer: OK.
Person: Grasp the pyramid.
Computer: I don’t understand which pyramid you mean.
Person (changing their mind): Find a block which is taller than the one you are holding and put it into the box.

Just wondering whether you’ve seen my personal experience with Simon, the robot. Simon is learning, it’s not at all about NLP, actually the NLP is quite bad, but the learning algortihms are quite OK and certainly moving in this direction.

@Victor: what’s SNAI?

Erwin, Actually, I thought the learning wasn’t really happening at all.  I thought it would have deduced that the last object (yellow) went in bin 1, but no dice :(  He said, “I think this goes in bin 1 or in bin 2 or in bin 3 or in bin 4 or in bin 5 or in bin 6”

Don’t you think NLP would be more effective.. “Simon, all yellow objects go in bin 1, except ones with rounded edges, which go in bin 4”.  Then give Simon a yellow object with round edges, but with one jagged edge”, He puts into bin 4” You say, Simon, that was incorrect, he asks why, you say because it has jagged edges.  Simon then asks “I see, so :

all yellow objects go in bin 1, except ones with rounded edges, which go in bin 4, but if it has jagged edges, then where does it go ?

....etc..

It is not difficult to keep stats on how many times you are told that this color ball goes in this bin #, and predict the next, that’s not learning, not learning semantically anyway.  It is an entirely different thing to comprehend a discussion made up of complex NL statements….statistics is NOT semantics.

  You can perform a statistical analysis on color-to-bin# correlations, but statistical NLP, but these approaches won’t allow us to write software that learns via NLP. 

Anyway…

SNAI -stands for:

    SNAI

    Not

    Artificial

    Intelligence

LOL recursive acronym, Dave will love it ... it basically means I want to avoid the whole philsophical disucssions about AI, because its a joke, nobody really knows what intelligence or thinking is.  Deep, protracted philosophical debates about machine intelligence are a waste of time and have produced nothing and hindered research, we need practical, focused empirical research.  I am developing a PRACTICAL, USEFUL application that speaks and learns things via NLP….  and the debate about whether it is AI, or whatever else, is not my concern, it is irrelevant.. it is what the bot WILL DO, that counts. 

I got the idea for SNAI from GNU . ... GNU stands for GNU Not Unix

Andrew - yes, the block world was very limited, but not a bad start given the time period.  That program was hard coded yes, and did not understand the semantics of the sentence.

The doctor program example given above will be an excellent demo for my engine… but I was also thinking of examples of ordinary conversations as well… “Hi how are you?”. . .etc.  Those type of things are important also.

What I am asking is, give me some sample conversations, ‘dream conversations’ that you wish you could have with a computer program now, that you would consider it to be fully understanding and learning.

Good examples are Google with NLP, clicking ‘Help’ in Microsoft word, and instead of being frustrated typing a keyword and getting 1,000 results that you have to sift through, you can ask it in NLP what you want to do.

These are real world, productive applications.

Of course your examples can be just a casual conversation which demonstrates FULL TEXT UNDERSTANDING and SEMANTICS.

Victor Shulist - Sep 15, 2010:

...

Anyway…

SNAI -stands for:

    SNAI

    Not

    Artificial

    Intelligence

LOL recursive acronym, Dave will love it ...

Picking on me already? I haven’t even had my coffee yet!

Just kidding. Actually, I don’t have any real issues with recursive acronyms. I’ve more or less given up on such things, since the computing world is full of them. After, all, who am I to fight with such popularity? 

{Pardon me a moment… Taking notes… Remembered the second thing I was going to tell Erwin… New feature request in a moment!}

{edit} Just noticed another recursive acronym! CRS = CRS Really Sucks! {/edit}