comScore
Uncategorized Monday, August 29th 2011 at 9:05 am

Two Chatbots Talk With Each Other, Awkward Hilarity Ensues

By ( ) Comments

As many of you are likely aware, a chatbot is a computer program designed to emulate a human in a conversation. They have no other goal than to generate natural responses, and are sometimes used to attempt a Turing Test where a computer successfully tricks a human into thinking that it is a human as well. That’s all well and good, but some folks over at the Cornell Creative Machines Lab wondered what would happen when you let two computers designed to sound human talk with each other.

At first, the results seem nothing more than awkwardly funny. The two bots snipe at each other, throw out a few non sequiturs, and generally fumble their way through the conversation. Without the anchor of a human to talk to, the bots seem lost and adrift, desperately looking for something on which to cling. That’s when it got a little spooky for me since I realized that in thinking in those terms I was anthropomorphizing these talking AI programs. Furthermore, I noticed that I was assigning personalities to them: The “man” on the left seemed arrogant and smarmy, while the “woman” on the right seemed snippy and defensive. Maybe these chatbots are better than I thought.

Read on below for a video of AI conversational feedback.

(via IEEE Spectrum)

Filed Under | • • •
  • Anonymous

     http://tinyurl.com/3tnmjr8

  • http://twitter.com/wcdarling Wendy Darling

    I want to know what bot logic leads the male chatbot to declare “I am not a robot. I am a unicorn!”

  • Pukin Gnome

    The gay union of logic and flamboyancy…

  • Anonymous

    i cant believe this!! me and my sister just got two i-pads for $42.77 each and a $50 amazon card for $9. the stores want to keep this a secret and they dont tell you. go here, EgoWìn.com

  • Nordan

    Speaking of computers trying to pass themselves off as humans…

  • josh w

    Ummmm, I call BS. That wasn’t AI. That was just a chatbot program that was reading a pre-entered response.

  • http://twitter.com/MPelletierProg Martin Pelletier

    Random response pointing to a preset sentence. These chatbots are supposedly what will pass the Turing test even though the sentences are not constructed by the “AI”. The snide remarks are all scripted human speak.

    Chatbots are to the Turing test what trolls are to discussion. They’re an intrusion, and everybody gives them far too much credit for what they’re worth.

  • http://twitter.com/PBlisster Phil Bliss

    Totally agree Martin Pelletier

  • Luke T

    It was a scientific experiment, you can look it up. What makes you think it wasn’t AI?

  • Anonymouse

    i love the ending

  • Asdf

    Look up “Robot Unicorn Attack”

  • yar

    u mad bro?

    Oh crud, I just used a scripted phrase someone else wrote and I’m human.. we’re doomed !?!

  • Zero

    Not impressed

  • Michael K

    Always a lovely day somewhere.

  • Epsilon Ultra

    Well, it is because of the “malicious” users who speak to it strangely.

    Cleverbot, in my mind reacts to a sentence by analyzing it and the previous conversation and semi-randomly picks an answer.

    everything you say to cleverbot changes it slightly, especially your reactions to phrases few others have reacted to.

  • http://www.facebook.com/profile.php?id=100001553728932 Jack Marrinson

    Best way to drive an artificial intelligence completely insane: allow it to commune with the internet.

  • Robcobcroft

    I really don’t think clever bot learns anything based on what people say to it, it has no short term memory whatsoever, what it has is a database filled with words and phrases from which it selects what appears to be an appropriate response based on various keywords that get directed at it. Try telling it something simple like your favorite anything. It will forget it in 2 statements time. Try typing in key words only for instance: God God Sex God Female. and then repeat you will start to notice patterns in the response system.

    This ‘A.I’ is nothing more than a sophisticated parrot or mimicker of human conversation it has no actual learning or thinking capability of its own.

  • http://profile.yahoo.com/2LIWBNLR45R4EAXIYIAJD44XCU Random

    Completely agree. They are clever little things but they’re a sideshow, and I don’t think they can advance AI much. I don’t understand how they get apparently good marks in some Turing tests, because when I talk to them, even when I’m not deliberately trying to trick them, they seem to say silly things often and don’t seem human. Apparently when Turing test judges are actually told to be rigorous judges they all fail as expected.

    I think the problem is they’re ultimately just clever programs for manipulating strings, which means they will never be able to track topics throughout a conversation or give creative or introspective replies – or even check if what they say makes sense. For a computer that could truly pass the Turing test you would need a supercomputer project worked on by a large team of experts in various fields, it would need to run thousands of routines in parallel and the vast majority of the work, I suspect, would be far below text manipulation, in huge networks of concepts, similar to how we think the brain understands. I think it would be a good thing for IBM to try for their next big computing project, after Deep Blue at Chess and Watson at Jeopardy.

  • Lol

    how……..
    where can i get theat program where they talk with sound

  • http://www.facebook.com/beverly.glover.315 Beverly Glover

    According to them, themselves, they are Clever Butts.

  • Nizz

    “Don’t you want to have a body?!”
    “Sure.”
    “Au revoir.”

  • ducks

    lol