Self-Recognition seems to be an item that designers are trying to integrate into artificial intelligence. Is there a generally recognized method of doing this in a machine, and how would one test the capacity - as in a Turing-Test?
$\begingroup$
$\endgroup$
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
1
Interesting question. I don't think anybody knows a definite answer, but some rough-sketch ideas seem apparent. Think about what it means to you to be "self aware". You'll probably cite the way you "hear" your own thoughts in your head when you think about something. One can speculate that inside the brain, the various centers that are responsible for hearing, vision, logic, etc. are connected so that as you form a thought, it's being "heard" by the hearing regions, even though it's purely internal instead of actual sound received at the ear.
So in AI terms, it seems likely that self-awareness will somehow involve taking the "thoughts" formed within the AI, and feeding them back into the AI so that it "hears" (or, more broadly, "senses") itself think.
There's this weirdly recursive aspect to all of this, which - interestingly enough - is something Douglas Hofstadter talked about a lot in some of his book, especially GEB. He was probably onto something.
-
1$\begingroup$ The furthest that Hofstader's group developed this (not including ongoing work there) was Metacat - a self-watching version of CopyCat: science.slc.edu/~jmarshall/metacat $\endgroup$ Commented Sep 9, 2016 at 16:11
$\begingroup$
$\endgroup$
I think consciousness is mostly an attention selection mechanism. It also serves as a memory/reality lookup mechanism as well as a storage mechanism.
A salient signal will be detected, causing the attention mechanism to focus on the signal, bringing up more details of that signal from both reality and memory, at the same time. That very act of focusing and bringing those signals into attention causes those to be stored in memory too.
The stronger the emotional signals are that accompany that original signal, the more strength with which that memory will be stored. Later memory lookups of that signal will bring back similar emotions.
When we "focus on our own consciousness," like mindcrime said, we recall the same words we just uttered because as we say them they are being stored which we then restore with associated emotional context. The conscious experience is what it is like to utter those words, hear them, feel their emotional context, and then feel an emotional response to that context - and then to repeat that process iteratively many times a second. That's how self-recognition works in humans, I think. And I think animals do the same thing, just without the words - only emotions.
