Author: Vincent Vega
Date: 16:38:36 03/14/01
Go up one level in this thread
On March 14, 2001 at 03:48:00, Ricardo Gibert wrote: >On March 14, 2001 at 02:26:57, Vincent Vega wrote: > >>On March 13, 2001 at 23:20:55, Ricardo Gibert wrote: >> >>>On March 13, 2001 at 22:33:46, Vincent Vega wrote: >>> >>>>On March 13, 2001 at 21:11:45, Ricardo Gibert wrote: >>>> >>>>>On March 13, 2001 at 19:06:27, HECTOR MUNOZ wrote: >>>>> >>>>>>There are some who might argue that a computer chess program is not a >>>>>>demonstration of intelligence in particular, a program which uses Shannon's >>>>>>Type A Approach. I need to present a solid argument that such a program >>>>>>does involve intelligence. >>>>> >>>>>Trying to come up with "a solid argument that such a program does involve >>>>>intelligence" is too defensive and probably impossible anyway. If you want to >>>>>win a debate, don't ask any questions he expects and shift attention away from >>>>>difficult questions he might ask by answering questions with questions, etc. >>>>> >>>>>For instance, at the earliest possible moment, when he asks you to provide proof >>>>>that machines are intelligent, you counter by asking *him* to provide proof that >>>>>*he* is intelligent. The list of intellectual accomplishments the average >>>>>individual has is embarrassingly short, so whatever he says in response, you can >>>>>belittle and attack ruthlessly. Whatever he says, you can discount by pointing >>>>>out that it has been done before and so he is just "copying". Not a sign of >>>>>intelligence at all, etc. He will find it very embarassing. For most people, the >>>>>list is empty. Your implicit stategy then is: a program is at least as >>>>>intelligent as a person, since the average person can provide little evidence of >>>>>intelligence that is not easily disputable. >>>> >>>>Hmmm, I can find quite a few "accomplishments" that could be defined as >>>>requiring intelligence (because they require learning, adapting to new >>>>situations, generalization or reasoning) and can be achieved by most people but >>>>not by a currently existing computer program: getting a "human-range" score on >>>>an IQ test, learning a foreign language, driving a car without crashing, playing >>>>a team sport, summarizing an article, writing a computer program based on >>>>requirements... >>> >>>The only one you mention that is interesting is "summarizing an article". >>>Programs have been written to pass a standardized IQ test, >> >>OK, prove it. What is the name of the program passing WAIS or Stanford-Binet IQ >>test? Show me the references. > >I remember reading in Scientific American about computer programs being able to >do a subset of the questions on an IQ test better than humans. One example was >to guess the next number in a sequence. > >Don't forget, the OP only has to come up with a strong argument that machines >are intelligent. He does not have to show they are just as intelligent. Not quite. These were examples of things an average human could do that a machine couldn't. If he used your strategy of discounting his opponent's accomplishments, he would be have to show that none of these accomplishments requires intelligence (this could be possibly done by defining intelligence as strictly as possible but it doesn't seem very convincing) or that a machine can do these tasks as well as a human. Showing that a machine is capable of doing some miniscule parts of these tasks doesn't show that it is "at least as intelligent as a person." So I'm afraid this strategy would fail. You correctly say "the OP only has to come up with a strong argument that machines are intelligent" but your strategy doesn’t follow this approach at all, instead it places much higher burden of proof on Hector, making him show that a machine conforms to human standards of intelligence. My examples show why this strategy is not a very good idea; they aren't meant to help him prove his original case. >>>a parrot can learn a >>>foreign language, >> >>Repeating a sentence isn't the same as knowing the language. The facts are that >>no computer can come close to a human in translating from one language to >>another. Did you ever try babelfish or Systran? They will badly mangle any >>non-trival text. There has been a lot of research done in this area and the >>results don't even come close to matching humans. You can read any article on >>the subject to confirm this. So you are wrong on this point. > >Again, they don't have to do it as well. They just have to do it better than >random. They make mistakes, we make mistakes. How am I wrong? You are wrong because you called these examples uninteresting, implying that a machine could also do them. Clearly, this is not the case. I'd love to see you try to argue to researchers working on machine translation that they are doing uninteresting stuff because doing better than random is good enough to show intelligence. Somehow I don't think they would agree. > >> >>>on the Santa Monica freeway in California there is a >>>experimental program underway where the car is driven by a computer (besides, >>>the human record on not crashing is hardly stellar), >> >>You don't seem to be aware that this involved a platoon of cars moving one after >>another in a special lane without any other traffic. Much closer to a train >>than to the real highway driving. And going from highway to city driving would >>require another great leap. Again, the research shows that no computer is even >>close to doing this. So you are wrong on this point too. > >Again, they don't have to do it as well. They just have to do it better than >random. They make mistakes, we make mistakes. How am I wrong? According to this logic I can fly without any outside help. Technically, what is flying if not moving through the air? I can do that by simply jumping up. It will last just for a fraction of a second but like you said, I move through the air, a bird moves through the air, so we both can fly. The difference is in numbers. A bird can stay in the air orders of magnitude longer than I could and I can drive a car though the city with orders of magnitude less problems than a computer currently could. It is this exactly this difference that makes me wrong when I say that I can fly and makes you wrong when you say that a computer can drive. > >> >>>robo-soccer is a team sport >>>where robots take part and as for writing computer programs, >> >>With all due respect to the creators of these cool programs, robo-soccer players >>still suck compared to humans. I know, I've seen them in action. If you don't >>believe me, see the official RoboCup page at http://www.robocup.org/. Very >>first sentence on this page is "- By the year 2050, develop a team of fully >>autonomous humanoid robot that can win against the human world soccer champions. >>- " Notice the date there? So you're wrong on this one too. > >Again, they don't have to do it as well. They just have to do it better than >random. They make mistakes, we make mistakes. How am I wrong? Same problem as before, they _do_ have to do it as well if you want to show that they are as intelligent as an average human being. > >> >>>a non-trivial >>>program that does not contain bugs is pretty much unheard of. >> >>But humans, unlike computers, at least can create such programs, even if they >>don't work right 100% of the time. That's a _huge_ difference. >> >>>It is pretty >>>difficult to defend the behavior of humans as "intelligent" when they make so >>>many horrific mistakes such as undemine the environment, World War I & II, etc. >>>Human behavior does not bear close scrutiny. For the most part, humans act with >>>stupidity rather than intelligence. We make too many mistakes. >>> >>>Besides, it is the average individual that has difficulty proving their >>>intelligence. For Albert Einstein, it's no problem. My point was machines seem >>>to be no less intelligent than the average human being especially given human >>>folly. >> >>Umm, without getting into some philosophical discussion, doing unintelligent >>things doesn't mean that one isn't capable of doing intelligent things as well. >>Getting back to the original question, I simply showed how your debate strategy >>of arguing that an average human doesn't have "intelligent" accomplishments that >>a machine has could be rebutted with some rather obvious examples. > >Excuse me, but you gave examples as signs of human intelligence and I countered >this with examples showing that you don't have to be human or intelligent to do >these things. I never claimed they did it as well. Sorry, this logic makes absolutely no sense. I don't see how you can argue that getting a score of 5 on an IQ test shows that one is "intelligent" or how completely changing the meaning of some text when translating it can be called an "intelligent translation" or how writing a program that works 0.01% of the time is a sure sign of intelligence. Note that I do not say that a computer has to do all those things to be intelligent. I am merely saying that it has to do those things and do them as well as a human if your strategy is to work. >I did not have to. If doing >these things are signs of intelligence and machines do them, then machines are >intelligent. > >If you do not agree, then you must agree that these examples are >*not* signs of intelligence. First, the machines can't do them. Getting a 5 on an IQ test is not getting a human-range result on an IQ test. Changing the meaning of a text doesn't show the ability to learn a language. Crashing your car every 5 minutes in a city doesn't show the ability to drive a car. Being hopeful that you can compete against humans in the year 2050 doesn't mean that you can play team sports now. Outputting "" as a summary of an article or as a computer program doesn't show than one can do either of these things. Second, these examples aren't meant to show that machines are or aren't intelligent. Comparing machines to humans makes it much harder to demonstrate AI. Like I said in the beginning, they are merely meant to show that your strategy will fail. > >Why get so bent out of shape over this. I agree that humans are intelligent and >machines are not, but this is hard to prove in an airtight fashion. Yes? What makes you think I'm getting bent out shape here? Is it because I'm trying to clearly show why you're wrong? Please don't project your feelings onto others.
This page took 0 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.