Slashdot Mirror

← Back to Stories (view on slashdot.org)

CIA Tested Primitive Chatbots For Interrogation In the 1980s

Posted by Soulskill on from the after-their-human-interrogators-couldn't-pass-the-turing-test dept.

New submitter ted_pikul writes: Newly declassified documents reveal that, 30 years ago, the CIA pitted one of its own agents against an artificial intelligence interrogator in an attempt to see whether or not the technology would be useful. The documents, written in 1983, describe a series of experimental tests (PDF) in which the CIA repeatedly interrogated its own agent using a primitive AI called Analiza. The intelligence on display in the transcript is clearly undeveloped, and seems to contain a mixed bag of predetermined threats made to goad interrogation subjects into spilling their secrets as well as open-ended lines of questioning.

3 of 65 comments (clear)

  1. Later versions by MrDoh! · · Score: 5, Funny

    Recent versions were tested on youtube comments but someone forgot to turn it off.

    --
    Waiting for an amusing sig.
  2. Codename? by Skevin · · Score: 3, Funny

    > and seems to contain a mixed bag of predetermined threats made to goad interrogation subjects into spilling their secrets

    "Now, Mr. Jones, if you do not tell me what we want to know, you may well have the pleasure of finding out why the first four letters of my code name is 'Anal'."

    --
    "Twice half-assed makes an ass whole." --Solomon K. Chang
  3. Re:Skynet Joke by durrr · · Score: 4, Funny

    "Do you admit that Oh Please Stop Drilling My Heelbones associates you to the plot to assassinate american diplomats?"
    "Tell me more about You Dumb Machine, I've Already Confessed. Is he one of your associates?"
    "You've already admitted that I Won't Tell You Anything, if you can tell us more about Beta_testentry_2_remove_this_before_field_testing we won't hurt your family"