Background
How does it work?
The Turing Test is performed by recruiting three sources: the machine, the control human, and the interrogator We can consider these A, B and C respectively. C knows that either A or B is a machine, while the other isn't. It is C's job to ask A and B questions or engage in a conversation with them to determine which is the human (B). A attempts to mimic what a human may respond with in an attempt to trick C into believing that A is the human and B is the machine. The test has no set questions or answers, it is simply a test to see if a human interrogator can determine which of two sources is the machine. The goal of the machine is not to know every answer to the questions the interrogator asks, but to respond in a way that a human may be expected to respond. This makes sense, since you wouldn't expect a human to be all-knowing. The original test guidelines produced by Turing said that if the machine could trick the interrogator at least 70% of the time, it passes.
How this relates to Computer Science
The machine used in this test is essential a program or collection of programs that work together to build responses based on key words that they interrogator uses in their questions. Basic AI is possible through user input and string analysis. For instance, a program could check if the string the user input contains certain words like "rain", "cloudy", "sunny", "hot", "cold", etc.. and respond in a way that may integrate weather, to show that it understands the meaning of those words. Implementation of the random operator could generate a random response among a list of responses to show that it is capable of expressing understanding in more than one way. For example, the program could be made to choose from a list of possible weather statements like "Tell me more about the weather.", "I hope it clears up tomorrow", etc.. This is a very rudimentary example, and would definitely not pass the Turing Test.
Sources:
https://en.wikipedia.org/wiki/Turing_test
http://www.biography.com/people/alan-turing-9512017
http://tinyurl.com/h4s67ev
Hey Ryan!
ReplyDeleteIt is interesting to think about how we could integrate other forms of responses into the Turing test. As far as I understand it, the test only evaluates textual description but what would happen if the question showed both the human and the machine a picture and asked them to recognize it. Then the machine would require coding in computer graphics and neural networks to be able to pass the test. I have a blog on computer graphics that may interest you.
Hey Ryan!
ReplyDeleteIt is interesting to think about how we could integrate other forms of responses into the Turing test. As far as I understand it, the test only evaluates textual description but what would happen if the question showed both the human and the machine a picture and asked them to recognize it. Then the machine would require coding in computer graphics and neural networks to be able to pass the test. I have a blog on computer graphics that may interest you.
Hi Ryan!
ReplyDeleteThe Turing Test is a very interesting point. It could be said as one of the basic ideas for Robotics and artificial intelligence, since Turing test is a model for describing the algorithm. The test aims to find out the computer's ability to think, which brings out the argument of whether a machine could have "intelligence" or not.
Shuzhi