Author: KarinsDad
Date: 10:05:27 03/28/00
Go up one level in this thread
On March 28, 2000 at 10:16:12, Graham Laight wrote: >People define intelligence in many different ways. In the linked article, >intelligence is defined as the ability to solve problems. > >I would define intelligence thus: the ability to find information which is >relevant to the presented situation. > >In the article, intelligence is stipulated as being the ability to solve "new" >problems. "Baloney", say I. NOBODY can solve new problems. People solve problems >(or components of problems (sub-problems)) by treating them as examples of >problem types which they already know how to solve (or where they know where to >get the relevant information). > >By my definition, it would (will?) be possible for a machine to clearly >demonstrate its superior intellect to a human. > >My opinion of the article is that it is a good discussion of when a single >computer will have as much computing power as the human brain. Those of us who >follow computer chess, however, know that copying the human brain is not the >most efficient way to create something which can demonstrate intelligent >behaviour. Possibly. But have you tried other alternatives in your chess program? Virtually everyone writes an opening book, a move generator, and a search engine since this is the easiest way to computationally come up with a reasonable move for a given position in a short period of time. But, it is not very intelligent. It is, however, relatively exhaustive. And note, I am not saying that people do not put intelligent shortcuts, pruning, and evaluation techniques into their programs. I'm saying that using exhaustive searches (relatively speaking) is not very intelligent; rather it is very easy to accomplish. What happens when you want to solve a new problem which has many magnitudes more variations than chess? How do you write a program to solve it? In the future, you may write a program that mimics human thought processes, you'll allow it to learn, and you'll put it's massive computational and cognitive abilites to work. > >-g > > Most creatures are intelligent to some extent. However, it is only recently that we have attributed attributes of real intelligence to creatures and not just reflexes or instincts. For example, an octopus which cannot open a bell jar to get to a crab and eat it, will immediately be able to do this task once it sees another octopus perform the task. This is called learning. In a general sense, we do not attribute certain types of intelligence to computers (e.g. musical sense, artistic sense). We relegate computers to those areas of intelligence that are easy for us to comprehend for them (i.e. programmed computations); similar to how we previously relegated a few areas of intelligence such as survival instincts or territorial instincts to animals. Hence, it is difficult to comprehend of a computer that is truly intelligent (as per our various definitions of human intelligence). But, consider the following. Mankind has gotten to his current technology level after approximately 50,000 years. In that time, perhaps 10 billion people have lived on the planet. So, it took 10 billion people working an average of 25 years each or about 500 trillion man-hours to get us to our current technology level. That's a lot of time and effort, even considering that a major portion of that effort was concentrating on survival. If a computer can get to the level of human computational power by the year 2020 and we can see by Moore's Law that a doubling of computational power occurs approximately every 18 months, then how long will it take before a computer is as computationally powerful as 8 billion people or smarter than the combined abilities of everyone on earth? Answer: 2^33 times smarter than one person or 33 * 18 months or 50 years or 2070. So, in about 70 years, a computer could be at the computational ability that it took us 50,000 years to achieve. Now, I realize that the article indicated that Moore's Law would start breaking down around the year 2007, but of course, the article also indicated that as the computers get smarter, they can be used to create enhancements for other smart computers, etc. There is a good chance that Moore's Law may decrease in rate temporarily (to a doubling every 3 years for example), but then swing back around to something like a doubling once per year or something. And, of course, this does not preclude improvements with multi-processor technology. The bottom line from my point of view is that computers will one day surpass the overall single human intelligence threshold just like they are surpassing the single human intelligence threshold in specific areas such as chess and mathematics today (in agreement with you). And, one day, they will surpass the intelligence of an entire community. And, one day, they will surpass the intelligence of an entire country. And, one day, they will surpass the intelligence of the entire world. It is inevitable. And in our shortsightedness of not even perceiving that animals are truly intelligent, it is likely that we will have problems conceiving that computers will one day have a consciousness similar to ours and have maybe 25 types of intelligence (as opposed to the 12? current types attributed today). It is inevitable. One day, they will no longer just be tools anymore. Today they beat us at chess. At the end of the next century, they may be beating us at life. And limiting our thinking to just computational ability and not cognitive ability is probably a mistake; just like it is a mistake to consider animals to have only instinctual abilities and not cognitive abilities. 15 years ago, nobody really conceived of the idea that we would be playing in quasi-virtual reality environments with people halfway around the world in real time. And that was only 15 years ago. How much more can be accomplished within a hundred years? > >On March 27, 2000 at 23:19:30, John Coffey wrote: > >>Intelligence is a collection of many different skills (some say there are 12 >>kinds of intelligence) that we ascribe to being intelligent. So what is Super >>Intellelligence? Is it being able to do human like tasks better than humans or >>something we don't even recognize? If the former is true then we already have >>super intelligence on a limited scale. True super intelligence will be the latter. >> >>My point is that pen and paper is a better memory device than most people. A >>slide rule can do math better than most people. The advancement of computer >>intelligence won't be something that suddenly overtakes us but instead it will >>be integrated with our daily lives and become a part of us. The real super >>intelligence will be us with our powerful new tools. Yes, for today. And for tomorrow. But maybe not in 50 to 100 years from now. KarinsDad :) >> >>John Coffey
This page took 0.01 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.