If you were asked to defeat the computer in the Turing Test (a test that can tell if a machines intelligence can communicate similar to a human) what types of questions would you ask it to reveal that it is actually a computer and why would the responses have to be given by a human being?
9 Answers | Add Yours
I would ask it questions about poetry and music and how it would respond to some famous texts or pieces of music. Surely one difference between humans and machines (we would think) would be our capacity to respond to such examples of the brilliance of human expression with an emotional reaction. Recognising that machines are incapable of such a response might be an interesting way to progress.
I think I'd ask the machine to explain why it is (or is not) moved by a piece of music. Aesthetic reactions are often extremely subjective, and responses to music are often especially difficult to analyze and explain. Presumably a machine would need to have emotions in order to respond to such a question.
I, also, agree that jokes are a good idea. I would also suggest posing questions which have to do with emotions. Given that emotions are feelings, and computers don't have feelings, questions like this would be a good identifier.
It seems a good distinguisher of machine from human would be to ask about motivation behind actions or comments, behaviors or answers. Motivation seems, thus far, to be restricted to human cognition--computers don't act from motives of altruism, enlightenment, concern, compassion, greed, avarice, etc.
Maybe another possible line of questioning could be based on reactions to natural phenomena. Asking the computer to describe its favorite aspect of watching a sunset or what part of a natural disaster such as a hurricane it considers to be the most tragic and why - this would be another route to asking the computer to attempt to interpret human emotions and feelings, which is the area in which artificial intelligence falls down.
I think jokes are a good idea. I would also keep asking it how it felt about something. For example, I might ask it how it would feel if it's sister died in a car crash. I might ask it how a certain song makes it feel. The answers should tell us if it's human or not. Computers can't feel.
I'd hate to have to prove I was human via #2's test. It's not easy to explain why a joke is funny, is it?
I would try to devise questions that are based on an understanding of human emotion. I'd try to somehow create scenarios in which the answer can only be found by understanding the dynamics of human relationships. I don't know if it is possible to program a computer to infer emotions from situations (without being told directly that the emotions exist).
The Turing Test (or "Imitation Game") was defined by British mathematician, cryptanalyst, and computer scientist Alan Turing (1912-1954) in 1950, where a human interrogator would ask questions of a machine and another human, but the responses would be textual, so there were no external clues as to which was the source of the answer (machine or human.) If the interrogator could not distinguish which had answered, it could be assumed that the computer responses had achieved a level of human intelligence. As #2 suggests, humor would be an excellent test to distinguish man and machine. Another might be to query favorites of something; by this line of questioning, the human interrogator could discern a pattern in the responses that a human answerer would have, but an automated response would not, these being chosen at random by the computer.
I'd find a series of jokes and ask my communicator to list them in order of funniness and explain their reasoning. Comedy/laughter is an extremely organic mental process.
I've been broke for a few months now and I hate it... I just want a little, y'know, just a little to afford those finer things in life I see other folks have -- you know, like food.
Could an artificial intelligence program unpack the nuances of language and culture and conclude that food is not contained in the group of things called 'the finer things of life' and could it then measure how humourous this juxtaposition is for a human mind? A difficult task.
We’ve answered 318,960 questions. We can answer yours, too.Ask a question