Current mood:	contemplative
Current music:	Texarkana (R.E.M.)

Artificial Pain
I caught myself meditating today on artificial intelligence and pain, and whether pain was "real" or not. I was thinking about those new systems that the military is using for recruiting and some airlines is using to streamline web customer support.

Part of constructing an AI is designing it how to learn and adapt, and training it. To start with, a good AI will have a somewhat arbitrary set of weights assigned to help it interpret data and stimulae (this is good, this is bad), as well as feedback to help decide if it made a good or bad choice.

How does this translate? At what point do we say an AI is sentient, and at what point do we call that feedback "pain"?

I could see how it would happen - one day, an AI analyst is working with an AI, when suddenly the machine informs the analyst, "It hurts." The analyst discovers that the AI has come to associate a wrong choice with pain. Is it feeling pain like we do?

The first thing I think is, "no, of course not, we feel biological pain". But think about it, what IS pain exactly? It's just feedback, electrical impulses managed by a chemical network that turns on and off switches in the brain. It's no more or less real than the electrical impulses that flow through a circuit board. We hurt physically because our body tells us we need to, and chemical unbalances aside, we hurt emotionally when we process/interpret negative data.

Can you imagine how painful it must be, therefore, to be an AI? In my mental excursion, I imagine the AI informing the programmer analyst how to adjust its own program so that it can learn in a gentler, less painful way...