To take part in discussions on talkSFU, please apply for membership (SFU email id required).
AI and Self Awareness
Currently there are a lot of projects researching AI. Most are geared towards the Turing Test. The object of which is simply to see if a computer can hold a conversation with a human without being recognised as a computer.
There are 2 principle paths being used, top down and bottom up. Top down means giving the computer a big database, like an encyclopeadia and a dictionary and a basic set of rules. The computer than 'learns' how to access information and put it together in a meaningful way. Bottom up means building a machine with the capacity to learn and teaching it.
Top down systems seem to be good for producing expert systems, intelligent search engines and that kind of thing. But not self aware.
The best of the bottom up devices are still in their infancy. It takes a few years to get any real intelligence from a child. Its going to be the same with these, probably longer because they dont have some of our inbuilt hard wiring.
What do the rest of you think, are self aware machines possible?
There are 2 principle paths being used, top down and bottom up. Top down means giving the computer a big database, like an encyclopeadia and a dictionary and a basic set of rules. The computer than 'learns' how to access information and put it together in a meaningful way. Bottom up means building a machine with the capacity to learn and teaching it.
Top down systems seem to be good for producing expert systems, intelligent search engines and that kind of thing. But not self aware.
The best of the bottom up devices are still in their infancy. It takes a few years to get any real intelligence from a child. Its going to be the same with these, probably longer because they dont have some of our inbuilt hard wiring.
What do the rest of you think, are self aware machines possible?
Comments
too much human/physical capital dedicated to these projects to fail over the long run
Self-awareness is very different and I actually doubt whether any computer will become self-aware in the absolutely human sense. The fundamental thing is being aware of being aware of being. It would be impossible to make a computer program for this ability. I sincerely doubt something such as self-awareness can be simulated in a computer program.
You say computers are limited to a set of rules that they have to follow. One could claim that humans are limited to a set of rules that they "have to follow" based on their genes and the physical makeup of their brains.
But a cyborg would be cool.
1) Turing test this only tests whether it simulates a human's actions, not whether it actually thinks like a human. This is nowhere near a measure for AI.
2) Machine vs. Biology this dichotomy rests on the false assumption that there is any difference of substance between a machine and our brains. The machines may not yet be as sophisticated, or as fast, or as capable, but I don't doubt that we will one day be able to simulate (note, not "emulate") our brains with machinery, or even surpass it (quantum computing). So it is entirely possible to have a non-biological computer which is in every other regard human.
3) computers follow rules, humans follow...nothing? no, we both follow rules. the difference is that our rules are far more sophisticated than the best software of the day. But just because we can't yet simulate our brain's complexities, doesn't mean we can never do so.
4) Human variation Is there ANY reason at all to believe that computer programs, sufficiently advanced, will not vary as much as humans do? No there isn't.
5) Self awareness this is one part of being AI, but certainly not the definitive component (I don't think there is one definitive component, btw), I (or anyone, for that matter) could write a "self-aware" program in some tens of lines, and definitely no more than 100 lines, right this moment. It's far from being AI, though.
Now to address OP, so far AI work have focused on observing human behaviour and then trying to write code that make the machines emulate that behaviour, with this approach we'll never get AI, we might get computers that act 100% like humans, including the illusion of free will, but it will still not be AI. The real way to program AI is to examine the very qualities that make US human, the way we learn, think, etc, and copying that onto machine, and letting them decide by themselves how their behaviour will turn out to be. It may be very different from our own, but it'll be truly a product of artificial intelligence, and not that of emulation.
For a computer to do the same, they would first have to have self-awareness. They would have to develop their reasons for their choices and also be able to choose to do otherwise (not by random).