One fascinating topic for me is A.I... ever since Deep Blue defeated Garry Kasparov at chess in approx. 1992, there has been frenetic research and activity in this field...to what? What is the ultimate goal of complete A.I.? Is it achieveable?
08-23-01, 11:31 AM
i sure it is "acheivable" but that doesn't mean that we will acheive it. I'm not saying we won't, but i'm not saying we will.
You’re in the wrong forum, we have a forum dedicated to such questions.
For the record Deep Blue did not use any AI techniques, but pure raw computing power. It was simply able to test a vast number of combinations of moves ahead and very simply select the move that gave the start of a good sequence. It was actually very crude. What were remarkable were the reactions and comments made by Kasparov. He had played against many computers and claimed he was able to easily recognize their very mechanistic approaches. He became very angry when playing against big blue because the machine appeared to be adopting strategic styles he had only ever encountered while playing against people, he strongly suspected that there were people helping big blue.
He was beaten not by AI or people just a faster than usual computer.
This fact sets the scene for true AI which need to model far more complex events than a simple game of chess, and for that there needs to be a greater increase in computing power.
It is estimated that to emulate the human brain we would need the equivalent of 100,000,000 x 1GHz Intel processors all functioning together. If Moore’s law continues to hold (computing power doubling each year) then we should have the computing power to emulate the human brain in about 20 years. During that time AI will continue to be developed and will become increasingly effective as more computing power becomes available.
But what is the goal? There isn’t one really apart from knowing that smarter machines tend to be more helpful than ones that are dumb. But if AI software becomes efficient then it should achieve the equivalence of the human brain at some point and then continue past that, at which point AI machines will become the dominant intelligence on the planet. What does that mean for us? If the AIs are benign then we should survive, if not then that is probably the end of the human race. Others say that we must adapt ourselves so that we can compete with the AIs, and that means various forms of augmentation or even as far as having our own brains scanned and uploaded into the processing units that are also used to house the AI software.
Hope that gives you some idea of what could or might happen. Whether it will happen as I describe no one really knows at this point.
Arguably, that may beget a Terminator style 'machines taking over the planet' scenario.
Maybe, that is one possibility. The period has been given the name "The Singularity". Meaning a short period of time of very rapid technological development during which super inteligence emerges.
If we are willing to augment and adapt ourselves to take advanatage of the technology then we have nothing to fear. It would just be a part of our evolution, an evolutionary leap again. It is called the transition to a post human form.
Penrose territory this, and whether one can make a program understand what it is doing etc......
and that is with emphasis upon understand#
10-21-01, 08:48 PM
Originally posted by Cris:<BR>It is estimated that to emulate the human brain we would need the equivalent of 100,000,000 x 1GHz Intel processors all functioning together.
This is an old figure based on equipment that has been present for a decade (Not necessary with 1Ghz Intel)
The reality though changes when you move this to the Quantum Computer realm, For instance if your able to duplicate the same frequencies as used in the 4.7 speed of light experiment involving Mozarts 40th Symphony.
The speed of travel was 81 Picaseconds (0.00000000081 seconds) this would mean a message that travels at 0.005 seconds could be compressed to travel at the above speed, and other messages too could add to it to fill the same travel time.
This added to Parallel processing would allow you to have a system that is quite capable of mapping the mind or rivaling it, as long as you have a team of programmers at hand to patch things.
10-23-01, 07:24 AM
Do you know what the speed of processing is for the human brain? Also, do you have an idea for the travel time of a neural impulse through the brain?
10-23-01, 06:27 PM
I wouldn't be able to answer those questions without looking at some source material, since it's been a long time since I've look at particular pieces of information.
I decided though that I would search for some information, and stumbled upon a site that I think will fit the bill (along with information through "An introduction to Neural Networks".)