Can AI systems be created that could recognize itself, and recognize intelligence in other systems, and make intelligent decisions about the other systems? Mankind seems to be making progress in self-recognition but I've not seen evidence of one system recognizing other systems and being able to compare it's own intelligence with other systems. How could this be accomplished?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
In the abstract, mechanisms for self-recognition (I personally prefer the phrase 'metacognition', since it carries fewer spurious associations) and recognition of intelligence in other systems can be considered to be pretty much equivalent.
In fact, both can be characterized in the standard percept/action framework: the task in both cases involves (however coarsely) classifying/predicting the behaviour of a black box system.
Such tasks can be universally characterized in terms of frameworks such as Solomonoff Induction or (more recently) AIXI.
