[Bragi]

Hello everybody,
I just made a new release of my neural network designer, which also contains (the beginnings of) a small chatbot demo, called AICI. I think (hope) that with this new version, I am finally reaching  the edges of usability. Although there is still plenty to do, I guess it's time to move the project into a new phase. So I am looking for a few brave souls willing to:

• do some peer reviewing: does it make sense to you? do you see any holes? What do you think is missing? Is there anything that needs better explaining?....
• play a bit with the designer, you know, do the beta thing
• perhaps already try and add some of your own stuff to AICI (I am putting some more info together on this one)

I have put up a little download section where you can find everything. If there is anybody interested in doing some beta testing, let me know, and I'll put up a public bug-tracker.

[TrueAndroids]

Sounds really cool Bragi. Nice to see so many artists, designers, developers, etc on this forum!

[Freddy]

Yes sounds really good to me too.  Also very complicated at first glance

Be sure to check out my post on blogging and hopefully we can help get you some more exposure.

[Bragi]

Also very complicated at first glance

Yes, the learning curve might be a bit steep. It's all seriously out of the box. I was hoping to alleviate some of that pain through a good UI. I'm also not expecting many people to ever build networks completely from scratch, as I did, but instead add to the already existing code (for instance, make aici, based on an example, copy files). Or take a few existing modules to create a new network (for instance, take the 'core functionality', a module to do calculus and a module for visual processing). People with a programming background should find many things familiar, like variables, assignments, conditional statements (aka if/do/while/for each). At it's core it's very similar to a tuned up stack machine.

[wgb14]

Hi I downloaded the chatbot but it says the brain file is missing....

[Bragi]

Ok, that's a problem. Hold on, I'll do an install on a clean machine. I do know that it works if you also install the designer, since this also contains the AICI network. 
Did you manually start it?

[Bragi]

Found it: I don't have any spaces in file paths, you probably do, the installer tripped on this. Thanks for letting me know this. I'm updating now, if you download again in, say , 15 minutes, it should be fixed.
PS: you'll probably have the same problem with the designer, when opening a demo project from the start menu.

[Freddy]

How long have you been working on this Bragi ?  It looks really advanced to me, in more ways than one.

[Bragi]

A little over 2 years now. Yes the whole thing is fairly advanced. It's a bit a chicken and egg thing: even though the basic concept is very simple (give the link between 2 neurons a meaning), to get something working, you need to provide a whole lot of initial data, in complex structures. I kind of figured I'd better try to make the barrier as low as I could, and working with long lists of numbers didn't seem to meet that, hence the designer.

[Freddy]

I see, thanks for explaining.  Looks like you have done a lot in two years then.    I'll find some time over the weekend to explore more fully.

[lrh9]

Is it open source?

[Bragi]

The designer data is indeed included with the aici network.

[sunama]

I just read the last few posts of your blog.

What interested me most was your last post, involving this pic:



Have you managed to get the correct answer yet?

Have you tried using easier input:

EG:
User: I have 2 hands.
User: Each of my hands, has 5 fingers.
User: How many fingers do have on 2 hands?

This is more simpler and would be a better start.

Also have you tried putting in some basic mathematics.
EG:
User: what is two plus two.

What output do you get for this?

I think it will be very impressive if you can get the correct answer to the questions you asked in the image (from your blog).

[Bragi]

Have you managed to get the correct answer yet?

Almost, I bumped into a small delay this weekend (a bug-fix caused the core to run a little faster, revealing a few new deadlock situations, which had to be fixed to get back to testing). This week, the bugs should be removed from the count algorithm if there are no more unexpected issues. Off course, I will than have to move the search code over to the noun resolvers, which don't have the fancy search yet.

Have you tried using easier input:

EG:
User: I have 2 hands.
User: Each of my hands, has 5 fingers.
User: How many fingers do have on 2 hands?

Less statements are not equal to easier for AICI. It doesn't matter how many or few you give for the algorithm. What's important is the structural complexity of the statements, their order, completeness and if it can be learned ('on' is not yet handled correctly by aici and this is a word who's meaning can't be learned, it needs to be coded) . Also, I wanted to test a few more (hidden) situations that aren't included in your statements. Here's a detailed description of each statement:
- a human's components are bodyparts: AICI doesn't know the word bodypart. With this statement, we make bodypart behave like 'component. Also, this is used to check if AICI can find plurals in unknown words.
- a hand is a bodypart: the word 'hand' was unknown.
-fingers are bodyparts: finger is unknown. This is also to test if the system correctly cancels out the 2 undefined plurals
-a hand has 5 fingers, a human has 2 hands: the actual data that needs to be counted. this tests if it is stored correctly
-i am a human: this statement is very important: it binds the concrete 'me' to the previous abstract knowledge. It is used to test if  AICI can correctly link abstract with concrete. In your example, there is a direct link between 'i' and hands. With these statements, you don't have this, which is an added layer of complexity that I wanted to test.

Also have you tried putting in some basic mathematics.
EG:
User: what is two plus two.

This is actually easier to implement compared to the counter. The network has mathematical instructions (+-/*%), though AICI doesn't yet know the word 'plus'. Once it know's this, it is simply a matter of attaching the 'add' instructions.

[sunama]

Less statements are not equal to easier for AICI. It doesn't matter how many or few you give for the algorithm. What's important is the structural complexity of the statements, their order, completeness and if it can be learned ('on' is not yet handled correctly by aici and this is a word who's meaning can't be learned, it needs to be coded) .

Ok, Lets remove the 'on' and try something more simple.
What happens when you type the following:

User: I have 2 hands.
User: Each of my hands, has 5 fingers.
User: How many fingers do have?

I'm interested in the result of this input as I'm now thinking that I wish to incorporate this sort of "thinking" in my own program.

I must confess that although I have read your thread, it was difficult for me to understand what is actually going on. But, its pretty obvious that if you are able to get correct output from your program, then you are on the right track. What will be impressive is if you can get the correct output to the questions you gave in the image, you posted in your blog. That will be a breakthrough.