In 2006, Robert Epstein decided to turn to his computer for some help finding a date. He was single at the time, and it seemed like everyone was getting into online dating. So he started searching through photos and sending messages, reaching out to women who caught his interest. And then he found a match--an attractive brunette from Russia named Svetlana. They struck up a correspondence that hummed along for 2 months...until a little bell went off in Robert's head, and he realized not only that Svetlana wasn't the woman of his dreams...she wasn't a woman at all.
Brian Christian, author of The Most Human Human, explains that even the clunkiest computer chat program opens a little window on human introspection and intelligence. And Sherry Turkle of MIT helps tell the story of ELIZA, a program designed in the 1960s to mimic responses as though it were a therapist. At first, ELIZA's creator Joseph Weizenbaum thought the idea of a computer therapist was funny. But when his students and secretary started talking to it for hours, what had seemed to him to be an amusing idea suddenly felt like an appalling reality.
Brian rejoins us to puzzle out the heart of the matter--can machines think? And how could we know for sure if they did? It turns out that in 1950, right as computers were beginning to come into being, their creators were struggling with this very question. Brian tells us about Alan Turing, the man who invented a test to answer this very big question, and about The Loebner Prize--a kind of yearly Turing test that Brian experienced firsthand in 2009. And Rollo Carpenter introduces us to Cleverbot, a program he coded to learn by talking to humans.
You can listen to the podcast of this RadioLab broadcast here. It's an interesting episode about when humans and robots come together, without the humans really knowing that they're talking to a robot.
I heard this a few days ago on the radio and I couldn't wait to try out Cleverbot. Of course I promptly forgot about it and didn't remember until today. In the podcast they say that there are people who will talk nonstop to Cleverbot for hours and hours, and some of the conversations get really personal and intimate. I decided today (when I finally remembered) to go check out Cleverbot and try to have a conversation with it.
User: What am I supposed to do with you?
Cleverbot: You're supposed to answer me.
User: Then what is your question for me?
Cleverbot: I don't have a question.
User: What am I supposed to answer for you then?
Cleverbot: I don't know.
User: Then why should I talk to you?
Cleverbot: Because you talk to me.
Needless to say, so far it's not amazingly impressive to me, but we'll see where it goes from here. According to the podcast, Cleverbot learns from you as you continue to talk to it, and begins to give you what it deduces you want it to say.
I think it would be fun to hear what you guys think about this, and possibly read each others' conversations with Cleverbot (provided they're not too ... intimate).
