Spurred by the book I'm reading, (
The Cambridge Quintet)as well as the one I just finished reading (
Goedel Escher Bach), I've been giving a lot of thought to some of the concepts surrounding AI.
Currently, it's language that I'm pondering. The nature of language, the relationship between language and how we define intelligence, and the dependence of language on sensory interaction with "the world".
Here are a few lines of thought I've got swimming around right now:
* Is communication a necessary component of something that we'd be willing to accept as a machine inteligence? I'm inclined to say no, but with the caveat that the
ability to communicate is. Though whether it's a form of communication that we can comprehend seems unimportant.
* Is there a component to language that transcends description. Specifically, is there some ephemeral relationship between a word and the thing or concept it represents in the real world which we are unable to describe using language and therefore cannot be coded in any formal way. From a Goedelian point of view (Goedel proved that any formal mathematical system with sufficient complexity cannot be 100% complete in its ability to describe true statements), this is almost certainly the case. But does that preclude a machine intelligence that truly understands communication?
* Can machine intelligence be "super intelligent" due to it's physical base's inherent computational abilities the way it's often depicted? Or will the only form of machinery which we would be willing to call true intelligence have to exist in such a heirarchical stack of complexity that, just as we have little to no visibility into the physical workings of our own mind, it would have no access to the underlying computational sytems of its own mind?
I'm sure to add more as I collect and organize my own thoughts.