Artificial Inteligence

[Nevin]

There are three rooms with a computer in each room. The computers are connected by a cable that runs through the walls. In room one is a human. In room two there is only the computer, programmed to respond as a human. You are in room three. What question, command or comment can you send to both parties to determine which one is not human? (Asking them to knock on the wall is cheating.) Is it in fact possible to tell which is the computer? If you cannot detemine which one is human, no matter what you write, does that mean the computer is in fact alive?

[jenstar]

That'll be the Turing Test, then. Is it possible to tell which is the computer? That depends on how cleverly the computer is programmed. For instance, all of the respondents to answerbank are computers, but you'd never guess it, would you?

[stoo_pid]

Is the computer connected to the internet?  If not you could ask what the weather is like outside.

[derbyram]

Questions that elicit answers that are based on emotions and opinions rather than facts are a good place to start, like when was the last time you felt truly happy and why? but as <STRONG>jenstar</STRONG> said, it depends on the quality of the software as to whether or not you would be able to determine which was the computer. AI tests with fairly un-sophisticated software have been able to keep a conversation going for quite a long period of time before the human has realised that the software is just using set responses to key words and phrases.

[NoGood Boyo]

Ask each of them to define female logic. Only the computer will attempt it.

[Marvin84]

NoGood Boyo has a great idea... except.... instead of female logic.... howabout explaining the "logic" of Windows....!! ;-)

[becks]

I'm woth derbyram a simple "how are you feeling?" would surely generate an answer which would be clear whether it came from a human or computer...

[ll_billym]

I reckon that you could do it very easily by asking something like the following:

"whta !s ur n4m�?"

The human would know what you mean, the computer wouldn't.

[morg_monster]

jenstar is right, it is the Turing test basically the test for AI, no computer has convinced a specialist they are really human. I think they have to ask them lots of questions though not just one.

[Nevin]

ll_billym is the closest for my money. Though I think a computer boffin could tell me a way round it using software. The thing that sprung to my mind was to somehow ask the respondents something designed to make them spontaneously rebel. Refusing to answer a personal question is not enough - it would have to involve inference that could only come form persoanl experience.
Or, you could ask them to try and guess how old the questioner was. The only clue would be in the questions they asked. Wouldn't the human always get much closer to the right age than computer. To take it further you could ask both parties to complete a profile of the questioner. Again, surely the human would always beat the machine, no matter how advanced the programme.

[Bliggin]

The problem with all the answers posted so far is nobody has really looked at the question. It is indeed philosophical in nature. Many people are getting bogged down in the idea of actual computer programming and actual AI. The question says that the computer would respond as a human would. This means that it has all the emotional, cognitive, and even spiritual capabilities that you or I would have. The real question here is, is the computer alive? Does the represention, a perfect or at least a believable one, make you indeed alive. The answer to that question is NO! To be alive really means to have bodily functions that work. To be alive is only physical in nature. Human representation does not equal alive. My dog is alive, he is sitting next to me wagging his tail, I fed him this morning and took him for a walk. Yet my dog does not exhibit any human qualities. I think the real question is, is the computer HUMAN?

[stevie21]

" If you cannot detemine which one is human, no matter what you write, does that mean the computer is in fact alive?"

If a man on the telephone puts on a convincing female voice, does it prove that he's had a sex change?

I think rather than "alive" you perhaps meant "intelligent" but even then, to answer your question, you'd have to define what "intelligence" is and psychologists have been unable to do this for decades. Good luck.

[Calvesy]

You could ask: "What would the other one say if I asked him to divide one by zero?"

 

[Nevin]

He might say it doesn't fit.

[Nevin]

I think I need to put the question in a wider context. The Turring test says, philosphically, if you can't tell the difference then the machine is alive. After all (this may sound ridiculous, but philosophically it's quite rational). If a machine had all the informaiton, and training that a human has (and larger memory and more complex software is making it possible), then it should be impossible to tell the difference between a robot and a human. In Blade Runner, Harrison Ford's testing machine would always fail. The point is there is no question that can help you differentiate. There is no question any of us can ask each other to prove we are alive.
However, somene else said this is all nonsense. They presented the Chinese room. In this case, we are asked to imagine a room, where Chinese symbols come through the door. We have to match those symbols to other symbols in a book on a table and send the new symbols back out the door. The point is, we have no idea what we are saying, like a computer has no idea what it is doing. It is simply sending back symbols it doesn't understand. It is not alive, not matter how alive it seems to us. The problem I have with this is, that if you think about it long enough, the same can be said of humans, can't it?

[peppyuk]

Just keep asking the same question, the human will get p***ed off after a short time.

[Diablod666]

as you said in your question, the computer is PROGRAMMED to respond as a human. This means it is not really alive, nor is really a human, because it is only doing what it has been programmed to do. humans and other living things cannot be programmed, they have free will. If it was possible to make a computer that had free will, and was not simply programmed to mimic it, i would still not call it a human, or alive. But i would call it equal.

[Diablod666]

another thing to think about is that all humans make mistakes... so if a computer was programmed to respond just like a human it would have to randomly make mistakes... this would be virtually impossible because it would have to categorise questions into how hard they are to answer, and then calculate the probability of the likelyhood it is for a human to get the answer wrong, and then it would have to somehow decide whether it was going to make a mistake or not, and then what would it put instead? something similar? also, computers cannot even decide, so there would have to be a cutoff point as to the probability that it would make a mistake at. The only way to get aruond this would be another randomisation :S

confusing

[Dagman]

Just ask "why?" the computer will top itsself trying to answer it! Saw this on an episode of "The Prisoner" where the hero outwitted a brainwashing computer with this question.