Lab 11 /

Which chatbots were the most convincing? What was done well? What done poorly? Within 10 years, will there be a program that can pass a five-minute Turing test? Is the Turing test a reasonable measure of whether a machine can think? More generally, if a program can behave in a way that is indistinguishable from human behavior, can it be said to think, and, if not, then what is thought?

As part of the requirements for Lab 11 of Informatics 102 students enrolled were to have a 5-minute chat with 3 chatbots and take a Turing test to decide it the 'bot' on the other end was a human or a machine. The three cyberbots were Eliza, Alice, and Jabberwacky. Eliza was programmed to model the speech patterns of a Rogerian psychologist. When talking with Eliza, circular logic becomes a pattern. It does not understand meaning of many words, which ends up creating sentences that make little sense. For example, because it is programmed to respond to the short response of "Yes/No" with "You are being positive/negative", when the user (in this example, me) replies with "I am not" (as a method of disagreeing) Eliza assumes "not" to be a state of mind (not as a method of disagreeing). I found Alice to be on the annoying end. Jabberwacky proved to be the most interesting of the three. The statements made were quirky, and a little more "human-like". For the most part, though, the statements seemed really confusing and not really fitting with the conversation.

The Turing test was most obviously a computer. I attempted speaking with the bot several times, and my internal compass never grasped a feeling of a human being on the other end. As stated in Shieber's article The Turing Test and the Loebner Prize, it was "realized early on that given the current state of the art, there was no chance that Turing's test, as originally defined, had the slightest chance of being passed by a computer program." So in order to give the computer a fighting chance at fooling any human performing a Turing Test, restrictions on the conversation were put into place. These limitations included two things: topic of the conversation and tenor (i.e. emotion).

I believe that if a cyberbot can achieve the intellectual conversation level of a human, it is the achievement of mankind that took it to that place. It is not 'thinking' until responses no longer need to be programmed and abstract connects can be made as swiftly as the human brain. I also believe that cyberbots have a place in our civilization. For example, as a method to supervise the Internet. I believe making the Internet safer for children and absent of visual stimuli that could be considered abusive (i.e. child porn) would be a valuable application of A.I. On this platform, the cyberbot could monitor chatrooms or help in catching online preditors.