Poe and the Turk

chess-playing turk

I was reading Poe’s “Maelzel’s Chessplayer” and came across a mention of a magician who seems to be the first chatbot. It answers questions given to it in the form of tokens. Anyone have suggestions for earlier ones? The Oracle of Delphi? The I Ching?

Well, maybe I’m getting sloppy about my categorization. First, we need a definition of a chatbot:

Working definition:

a mechanized conversation entity that responds to input

This would rule out the I Ching.

7 Responses to “Poe and the Turk”

  1. 1 Christy Dena

    It’s interesting how we separate the history of robots with chatbots. I guess it depends on what it is we’re fascinated by with chatbots: their response to specific things we type or say — to our thoughts — or to the environment (which can be rolling around an obstacle. The environmental skills and the ‘conversation’ or language parsing will ultimately converge. So, I like to look at automaton in general because the reactions are pretty similiar. Here is an software example of The Turk (the chess machine). I would still get a buzz out of reactions from the software even though our ‘conversation’ is not with words.

    But I guess we’re attracted to different elements of autonoma: and we happen to be interested in the ‘conversational’ aspect. I don’t say word output because I’m not as interested in story generating software as I am in the process of me inputting my thoughts and the bot responding to them. That for me, is one of the attractions to bots I guess — that they respond to your thoughts (the off-track stuff we’re trying to fix) unlike humans who respond to triggers in their own thoughts. As if talking with a bot means talking with an unbiased person…

    But back to the a definition. The textbook by Stuart Russell and Peter Norvig, Artificial Intelligence: A Modern Approach outlines

    [F]our basic kinds of agent program that embody the principles underlying almost all intelligent systems:

    • Simple relfex agents;
    • Model-based reflex agents;
    • Goal-based agents; and
    • Utility-based agents.

    The simpliest kind of agent is the simple reflex agent. These agents select actions on the basis of the current percept, ignoring the rest of the percept history.

    is also the aspect that you have touched on — that bots are channeling divine…that there is a force greater than human thought and so creating a ‘machine’ allows it to speak to us through it. There is, incidently, a great Oracle in the Winged Sandals site.

  2. 2 Mark Marino

    I’m interested in what you say about the bot being an unbiased conversant, one without its own agenda to get in the way.

    I’d like to hear how you see the environment and conversation converging.

    Incidentally, one of the Turk operators was able to guess its opponent by the way the person played. In other words, he was having a conversation with someone through chess, much as you suggested.

  3. 3 Mark Marino

    I wonder if we can’t add a few more features:

    Their speaking
    Their environment

    The context:
    This would speak to issues we’ve discussed before, the frame or persona of the bot, whether Jack the Ripper or a Turk (as in the case of the 18th Century automata). I think this is a crucial part of the framing of the agorhythm.

    What do you think?

  4. 4 Jeremy Douglass

    I’m revisiting this post, because I just acquired Tom Standage’s book “The Turk: The Life and Times of the Famous Eighteenth-Century Chess-Playing Machine.”

    Having skimmed the book (which looks to be a very interesting history) I have a few reactions to it and the post above:

    The Turk was not a chatbot, it was a chat client, albeit one driven mirrored chess boards, a magnet display, and a moving mechanical arm. In early showings, the Turk would perform like a Ouiji board and spell out answers to questions on a chessboard marked with letters - this was later dropped from the performance because it didn’t seem credible to educated audiences of the time, while chess did. So in general we could talk about how the token-magician was represented - as either an interface to some remote spiritual logic, or as an ingenious mechanism. Do we define a Magic 8-Ball as chatbot/not based on our beliefs, or based on how it works?

    In terms of how it actually works, the Turk was an interface (real human) that claimed to be a mechanism, whereas the magician was a mechanism (random token)… did it claimed to be an interface? The first interface between humans that claimed to represent non-human intelligence is probably somewhere in the history of religion - speaking statues, disappearing and replaced offerings, etc. etc. - but the Turk interests me because it might be the first interface that claimed to represent only a mechanism (rather than a spirit or god or spell etc.).

    Standage has an interesting comparison between the Turk and the contemporary battle of Kasparov v. Deep Blue, in which he compares some of the claims and stances about what makes something an authentic machine interlocutor vs. a human-in-a-cabinet. In particular, what comes out is some continuity between Alan Turing’s first paper-based chess program and the argument that Kasparov actually was playing against “a team of IBM researchers” - that is, that the later examples have likewise been attacked as hoaxes, albiet open ones, in which a Frankenstein-like assemblage of pre-existing bits of human intelligence is playing rather than the product of a generalized machine intelligence which learned to play chess. This argument (and Searle’s Chinese Room argument) has been made about AIML chatbots as well - that their pre-defined outputs at the level of the phrase and sentence disqualify them as ‘true’ chatbots.

    What I draw from this is that, as you said, defining our terms defines our answer. We might say that “The Turk, like modern chatbots, hid human intelligence within the black box of a total machine process” or we might say “what the Turk displaced in space (the human operator hid under the table) today’s stored-program interfaces displace in time (the programmer, who attempts to anticipate and react to his future human interactor)”….

  5. 5 Mark Marino


    If you enjoyed that book, you may also enjoy:
    Edison’s Eve : A Magical History of the Quest for Mechanical Life

    I’m interested in your analysis. You write: “the Turk was an interface (real human) that claimed to be a mechanism”

    Isn’t the Turk, along these lines, more like a puppet or an avatar than an interface that claims to be a mechanism, like an NPC in an online game who turns out to be a person? In this analysis, would the chess-board be the interface or the room (or text) in which one encounters the Turk?

  6. 6 Jeremy Douglass

    True, my language was imprecise. Avatar is a better word for that case.

    The Turk is a human avatar (PC) that claims to be a machine automaton (NPC). In both representation and reality I’d say that the physical Turk is itself an interface - although in one case, like a chat client (AIM) is an interface to another human, and in the other case like a chatbot client (AIML) is an interface to a program.

    Representation: The chessboard collects input via magnets (interface). The clockwork gears processes the input (code). The arm transmits the response (interface).

    Reality: The chessboard collects input via magnets (interface). The human reacts. The arm transmits the response (interface).

  1. 1 Amazon Mechanical Turk at WRT: Writer Response Theory

Leave a Reply

thesis writing service