I am diving into research to try to create strong AI. I am not a fan of Machine Learning, or at least what I have seen of it. Probabilistic methods that don't try to understand the complexity of the underlying problem but label it as "randomness" are setup to approach the problem wrong.
Yet, the question as to what human-equivalent intelligence would look like is almost impossible to answer. What is intelligence? I won't try to answer that here. In fact, as you can probably see if you're reading this on a large enough screen, this is a short article. So let's drop human-equivalent for human-comparable and forget about intelligence altogether. Instead, let's think about language.
I would like to build a discursive machine; one that converses with human enough methods that one has the feeling there is someone on the other end and that is interesting enough that it's statements develop with experience. What do I mean by "interesting"? I mean complex, reorganizing, patterned, etc., but the truth is I don't quite know yet. If I knew a rigorous answer to what I meant by "interesting", I would know where to begin looking for first steps. As it is I have a couple of hunches, and those are where I will begin.
It's as simple as this: I want something similar enough that I can empathize with it, complex enough that it seems to think, and alien enough that it runs on human-built technology.
Thus I begin my search.
No comments:
Post a Comment