I am willing to share everything but my source code.

Laura Patterson - Feb 9, 2012:

Victor, we are all here to share our ideas and hopefully together we can accomplish something great.
At least that is the case with most of us.

Some of us though are just here for the incredibly sexy fashion of wearing a black & white striped shirt, and carrying a whistle.

Laura,
Seems very promissing the context evaluation aproach you are using. Word association is fantastic because it is one of human instruments for learning. I know it’s not the same, but it can expand AI knowledge as it goe doing assocations… I think so…

Victor,
I loved the idea of a AI learning and teaching languages .
Just 2 questions:

1) Will the machine be able to interchange a knowledge basis learnt in a language to another?
Something like: GRACE learns something about sports in French but a person asked her about this subjetc but in English. Will GRACE be able to translate her knowledge base from French to English or would it be necessary to teach her again about the same subject but in French?
Confuse? I don’t know if I made myself clear.

2) For example if a user wants her to speack Spanish. Would be possible to a simple user to teach GRACE the Spanish grammar or would it be necessary a programer ( you) to do so?

Hi again Fatima

My intention, (and time will tell if it works this way, after all this is highly experimental at this point), is to keep the original statements entered in their original language.  Semantic markup added to say an input in English would correlate with semantic markup added to another, previous knowledge base article entered in say French.  I imagine I will probably have a “language specific driver” which will query the KB.    So input statement comes in in language X, semantic marked up M1 is added.  Then, later, QUESTION comes in, in language Y, but marked up M2 is added.  Now the actual internal codes of M1 and M2 will be in English (any NL language would do of course).

Actually the way Grace responds right now, is not based on the actual words, but properties that are added.

For example, you’ve noticed the way she finds the conjugate of verbs in the above examples right?  “Bob TOOK out his suit….” . . “Dib bob TAKE out his suit?”

actually, I don’t know if I included it in the above examples but she can also be told “Bob went to a celebration”...  then “Did bob go to a party?”  and she’ll relate the two because they both share a property, in this case “human-gathering”.

So by the same system, Grace doesn’t really care about the actual words, just relationships of their properties.  For other language, it is the same thing.  Relating “party” to “celebration” is the same as relating to words in different languages.

Now, right now she considers party and celebration to be the same because of that shared property… but , in other contexts she may be educated that other properties are important.  Some people are very picky and may not want to consider celebration to be same as party, and thats great, they can educate Grace that there is a difference if they like.  Other people, many words mean the same thing, even though they really don’t.  Synonyms seem to be a personal choice in a way.

So yes,the original statement will be stored, in its original language.  The comparison will be done always “at runtime” ..there will be no big conversation between all languages to a single language.

for question 2…..

I have created a mini-language for describing grammars.  PTSGL - parse tree semantics generation language.  So, for the foreseeable future, you’ll have to code the rules for a new language.

But yes, it is on the (1,000,000+  item “to-do” list) to have her learn via discussion.  She will for sure learn new terms and their meanings via conversation.

Actually, I am also working on a “GUI” version of grammar training.  Where you can just point and click to create new rules…. I guess something similar to what Jan is doing.

Oh My God!

Sorry! I simple lost the timing of this topic!
Statistics and my Tesis are killing me!!!! hahahahha

GRACE is really interesting.
I asked you about interchanging languages because I imagined the program in a situation it could talk to different people in different languages and be able to say what people just told. Like a translator, but at the same time with the ablility to store the new “knowledge”.
It would be something almost human. ( oh my imagination!)

But for sure it is just Sci Fi now, I imagine.. coz as you said in the current state of art AI is not able to understand the meaning of words.

I’d love to see an Asimo or a Nao/Romeo using GRACE’s or CR’s Alex abilities.

A robot with nice conversational abilities and able to interact with the real world . It would be amazing! And pretty more useful for aged and disabled people.

Fatima Pereira - Mar 27, 2012:

Oh My God!
Sorry! I simple lost the timing of this topic!
Statistics and my Tesis are killing me!!!! hahahahha

Funny you should say that, I am currently figuring out how statistics/probablility will factor into Grace’s ‘confidence formula’ for deciding which parse she choses   She does fine so far without it, but I don’t think it will hurt to integrate it in an ‘auxiliary’ or ‘tie breaker’ kind of way perhaps…. experiments will be done.

Fatima Pereira - Mar 27, 2012:

GRACE is really interesting.
I asked you about interchanging languages because I imagined the program in a situation it could talk to different people in different languages and be able to say what people just told. Like a translator, but at the same time with the ablility to store the new “knowledge”.
It would be something almost human. ( oh my imagination!)

Yes, translation -is- an option, but much lower priority.  I think Google has that pretty much conquered.  However, to be told something in English but asked a question on that by someone else in Spanish, that is still a promising area.

Fatima Pereira - Mar 27, 2012:

But for sure it is just Sci Fi now, I imagine.. coz as you said in the current state of art AI is not able to understand the meaning of words.

No digitial machine made of silicon will ever understand words the way a flesh and blood, electro-chemical mind does. 
However, Grace has a functional understanding of words and phrases that define them - how they relate to each other.  And yes, that ‘web’ never really touches physical reality…  I don’t think it will have to in order to create a very powerful and useful tool.

Fatima Pereira - Mar 27, 2012:

I’d love to see an Asimo or a Nao/Romeo using GRACE’s or CR’s Alex abilities.
A robot with nice conversational abilities and able to interact with the real world . It would be amazing! And pretty more useful for aged and disabled people.

Oh yes !  One of these days.  The right hardware, the right software, and the right data structures.