The hardest thing about artificial intelligence is being able to make a program sentient, or self-aware. To accomplish that, you need to have an algorithm that allows for self-monitoring, feedback, and thinking pipeline that is interruptible. The following is one type of AI that leads to self-awareness (the type you see in HAL and skynet) http://www.edepot.com/forums/viewtopic.php?f=12&t=1534 It goes into the details on a strategy for getting self-aware programs off the ground, and the requirements.

If one says that "self-awareness" can be programmed into a machine, one misunderstands the concept of "self-awareness".

We (science) don't understand what makes humans self-aware or conscious, and frankly we can't even define those terms scientifically. Heck, we can't even figure out if non-human animals are conscious or self-aware. Thinking that you can program something that you can neither understand nor recognize is pretty optimistic.

-me

I think we need to be carefull of the semantics of the term "self-awareness". If an AI agent is aware of its location, it's current states, etc., that is a form of self-awareness.

Ah, another troll that couldn't teach a monkey to eat bananas, but claims to have figured out the path to self-aware machines (whatever that means). Gotta love it.

Having the AI machine self discover itself is much more than being programmed with 'I think, therefore I am'. Its more than a discovery mechanism and a generalization/relational system with huge capacity and fast processing.
Its more than a sophisticated planning/conceptualizing system.


The machine would have to conceptualize what it really means (not just file it somewhere under 'I exist') and it would have to 'understand' the difference with what 'everything else' is. Such a fundamental concept/idea ties in with every
thing that the machine 'knew'.
Its whole data set suddenly has another dimension added to it (a likely spot for insanity to set in).

Have they traced the point where humans 'realize' themselves and is it more like switching on a mechanism that is already has the capacity/inclination or is it a gradual process (with stages??)

A few comments...

1. Prove to me (or anyone) that YOU are self aware. I'm not trying to be a dick, but unless you can prove that you are self aware, how will you prove that anything else is self aware (let alone define what self aware is).

2. Why? What is the advantage of self aware?

3. I once read that self-awareness is the minds way of running a simuluation of self. In this way, any program that incorporates it's future state in to its present equations would be made self aware; in otherwords, Windows XP is self aware, your TCP/IP stack is self aware, and so are most telephones. :)

Plus, it is not a good goal anyway. HAL and SkyNet were bad guys remember?

ironman + Skynet =?

Hi,
The god is everywhere for everyone's help. He takes care of every person and about his feelings.


Joe