AI "middleware" language?

[one_raven]

This mat sound silly.
Or maybe it has been thought of and someone is working towards it.

Anyway...
The other day I was thinking about communication with an AI device.
One of the biggest problems I see is language structure.
English, in particular, since it my native language.

Latin, however, or a similar language seems very structured and reliant on steadfast rules (I could be wrong, since I don't speak Latin) without exceptions or subjective interpretation.

If we create a language (or simply use Latin or another existing language) that strictly follows rules of verb conjugation, sentence structure, inflection of nouns etc. we could have a logical base language that could sit between the user and the interface as a form of "middleware" language.

Then, once the language is fully built, simply build "foreign" language interpreters to sit in front of it.

Is this making sense to anyone but me?
I don't feel like I am expressing my thoughts on this very well.

 

[G71]

one_raven,
check out the "Private AI development?" thread.

 

[one_raven]

Thanks

 

[G71]

one_raven: Check this out. The Webmind team reportedly developed a code which converts regular English to the KNOW (Knowledge Representation Language of Webmind). I have the latest (newer) version (ver. 2) of the "KNOW Specification" document, but it's not online at this point. I agree that this is something very important. It has often a major impact on how is the system's database designed (=key thing). I'm trying to design a communication module for my own AGI system and the KNOW specifications helped me a bit. I'm still working mainly on the theory. For now, I do not have a good idea how to convert regular English text into that AI-friendly language. I'm still learning English myself ;-)), but the conversion is definitely a pretty tricky thing. I'm REALLY curious how well it's handled by the Webmind's code. My system is gonna be a bit simpler. I'm not planning to work with regular English sentences on the input side (+ no emotions when thinking). It's gonna be English (I/O) but the communication is kind of Menu/Form based, which means the system cannot simply browse the full-of-garbage-Internet and learn from it, however, the system should know FOR SURE what is the user talking about.. It's just tricky for me to find a good balance between User-friendly and AI-friendly communication formats.

 

[wesmorris]

I don't think you're gonna get "true AI" as in a sentient being without the ability to abstract its experience into concepts which can be developed over time. If you can get a sentient being it could learn the concepts of the english language and you'd be fine.

Hmm.. can you have "intelligence" without being self-aware?

 

[G71]

wesmorris:

ability to abstract its experience into concepts

this IMHO can be done even when the AI app is usig just text I/O.

If you can get a sentient being it could learn the concepts of the english language and you'd be fine.

Too complicated (at least for me).. Requires lots of resources I currently do not have (and maybe no one has).

can you have "intelligence" without being self-aware?

Yes, but it's not a big problem to make an AI system self-aware.

 

[wesmorris]

Originally posted by G71
Yes, but it's not a big problem to make an AI system self-aware.

Please elaborate.

 

[G71]

wesmorris: Let's discuss this in the "Intelligence and Awareness" thread. See the link I posted there. You will find my definition of self-awareness which may give you an idea how to implement it.