Wednesday, November 19, 2008

Turing Test - Chatbots

The Turing test "is a proposal for a test of a machine's ability to demonstrate intelligence." [1] It was first introduced and described by Alan Turing in the 1950 paper "Computing Machinery and Intelligence" in which a form of examination takes place whereby one human and one machine engage in a conversation, each of which trying to appear human, and if the judge cannot reliably distinguish the machine from the human, the machine is said to have passed the test. This was definitely a paradigm shift at the time and surely it can rightfully be labeled as the question/test that prompt the notion of "computer intelligence" or, with a term we are so familiar with, "artificial intelligence."

Since the 50s, various alternations have been made to the test due to developed ideologies but throughout the history of the Turing tes there had been four major turning points: 1) the publication of "Computing Machinery and Intelligence" in 1950, 2) the announcment of Joseph Weizenbaum's ELIZA in 1966, 3) Kenneth Colby's creation of PARRY described in 1972, and 4) the Turning Colloquium in 1990.

ELIZA works by examining the user's input through identifying keywords in the sentence or question and once found, the system is designed to respond with either "a generic riposte or by repeating one of the earlier comments." [1] Meanwhile, Colby's PARRY has been described as "ELIZA with attitude" in that Colby intended andwished to model the behavior of a paranoid shizophrenic through the machine responses.

I must confess that these chatterbots, although well-designed, still contain glitches and abnormalities in their responses. Here's a sample and short conversation I had with the machine:
A: what's up?
B: I am chatting with clients on the internet.
A: what clients?
B: Is that a rhetorical question?
A: no.
B: Oh so must really want an answer.
A: yes.
B: Aw.

Honestly, I'm still confused and still wondering how I would respond to the "Aw." It seems like no matter how inricate designers model their machines; whether it's the language or the voice of communication, there is still something off about these machine answers. Perhaps it's due to my subjective perspective but I do truly believe that at our current rate of technological developments and innovations nothing's impossible. But sometimes, technical inabilities hinder the realization of philosophies. Lastly, I don't think the Turing test stands as a credible examination of a machine's human-ness. Even if a program can behave in a way that is indistinguishable from human behavior it still cannot be considered to speak and think on its own. In the "Chinese room" sense, computers essentially lack the ability to formulate new ideas and theories and ultimately the ability to "think". Likewise, "a symbol-processing machine like a computer can never be properly described as having a 'mind' or 'understanding'." [2] Recent studio films such as I, Robot or Stealth all attempt to visualize a world where potential evolution of artificial intelligence could doom the earth. Thankfully, the machines from the movie eventually confide in humans and the actors and actresses live happily ever after but imagine what the world could be like in 50 years time. Would artifical intelligent become a prevalent element in our lives? If they do, I certainly hope their response mechanism aren't Turing, ELIZA, or PARRY. Who knows.

No comments: