Maybe I'm wrong somewhere...
I tried thinking about an A.I. outside of a game world, and imagined it in an email context where it has to sell stuff through emails, by constructing a message of its own chosen words list from a word db.
It can get feedback by receiving email responses, which can be interpreted into 'energy' and the bigger the email, the more 'energy' it gets. But, this is still a fixed solution, because I'm telling it that the more emails it gets (and bigger) the more 'correct' its actions are. I'm not sure if this is the correct way to inform it of its success, and it will be the correct in the context of selling stuff online, eventually it will start constructing meaningful emails that actually sell stuff, but if you get it outside of this context it wont be able to perform with a different set of input/output info, say flying the RC helicopter from my last post, or at least I'm not sure it will still be able to work.

What do you think?

Now, to clarify, the email seller A.I. wont know its actually picking words out of a db and selling stuff to someone, it is just manipulating data, its environment, into getting more 'energetic' responses... The same goes for the RC example, it wont know what a rotor is, it'll just have input/output of dangerous or safe distance from the ground and its available controls and real energy (battery charging?, but this requires also location mapping)...

the real problem is , the computer only does what it is told , within the confinement of the small amount of pre programed or pre stored intelligence.

the only way I see a.i becoming intelligent , is to design on what role the pc/cpu essentially needs to be in the process , the computer needs to be only the middleware , the channel and link between the complete process , where the input and output is bi directionally linked to a bio engineered brain , a brain grown in a lab and then coupled up to a pc , the brain needs to be feed the visuals and awareness of the virtual environment , it needs to be fed with all senses , motion smell taste hearing seeing feeling , and the brain needs to be able to control the virtual actor , it needs to be able to control each bone and have a feedback about motion and balance and everything to make the brain believe the virtual is truly real , then and only then would there ever be true a.i .

true a.i needs awareness , and true input from something that is truly aware and alive.

a computer is only the middleware in the process like it always was and always will be until true awareness is ever developed in a form , it will always need a bio form of awareness and life.

science have been looking past what exactly the computer is , its the central link only , they need to develop the main " life-form / interacter" which interacts with the computer (interpreter /central link in th system of true a.i)

I suspect that somewhere in the future such would be developed , however weird and evil it may sound.

Whatever AI you try to make it will still need some kind of training, just like every living being does. You know that something hurts because it really does hurt, which is a feeling you want to avoid. At some point you may know what food you like a lot and which you don't because you like the taste or you hate it, which is also something you kinda just know, although this may change over time probably based on lots of different influences, which in theory could be simulated.

Everybody learned by trying and getting some kind of feedback and others sharing their experience. Also at some point it is possible to make connections like "thing x hurts and thing y looks similar so it probably hurts too".

I really don't think a neural net is far off from what you are talking about and might even be capable of that given a complex simulation of human life, which however won't be easy to create and teach...