My favorite concept in AI by far is the Turing Test. It basically tests the ability of a machine to respond to human conversation the same way a human would. An example of this test would be me speaking in an online chat room. I could be talking to either a real human being or a robot. Ideally, if I was chatting with a bot, I would not be able to tell it was a bot, and would instead believe I was talking to a human! Now I’ve talked to chatbots online before (it’s totally not because I have no friends) and although I am surprised by their relevant responses to my inquiries, I have yet to meet a robot I cannot stump. Most cannot yet “remember” anything from previous questions and use them in their future responses. The day will come… and hopefully, I am a part of that day… when a bot correctly answers the question, “What do you mean?” Turing claims that if a machine passes this test, then it can “think.”
I think the goal of NLP programming is to pass the Turing Test. This doesn’t just mean with online chatbots, although those too can have valuable applications such as therapist robots and intelligent help hotlines. Imagine if a robot could give me valuable life advice! After all, I only turn to robots when humans can no longer help my situation...so, all the time. Any instance where a robot is indistinguishable in its behavior from a human has reached the pinnacle of Artificial Intelligence.
Now here’s where it gets interesting. In response to Alan Turing’s Turing Test, John Searle came up with the Chinese Room argument, holding that a machine can never have a “mind” and can never “understand” human input, no matter how human-like it behaves! It states that a computer program only takes input (such as Chinese characters), passes it through a function, and returns a certain output (other Chinese characters). Even if this output passes the Turing Test, that is no indication that the program “understood” anything in the process or even “mimicked” the human brain. Searle claims that if he stood inside the Chinese Room with an “English version of the computer program” and sufficient writing supplies and received Chinese characters as input, he could easily return the correct Chinese characters as output. But he, knowing no Chinese, wouldn’t understand a single word of it. All he did was follow a step-by-step program.
This brings up the difference between Strong and Weak AI, Strong AI theories claim that a fully functional AI program can have the same sense as a human, and Weak AI theories claim that it never can. The controversy is very interesting! It is now 3 AM (yeah, I didn’t go to bed after that last blog post…) and I think I’ll finally get some shuteye. Now, make a decision, am I a bot writing my autobiography or a human simply obsessing over some pointless intelligence theories?
