Bots, Demons & Dolls

Demon Inuyasha‘Socrates had a bot.’ Well, what Andrew Leonard then goes on to say in his book Bots: The Origin of New Species, is that Socrates had a nonhuman companion: a daemon. Leonard’s book gives an alternate (from my perspective) view of the history of bots. His book, which I only just discovered a few months ago, was first published in 1997. I haven’t finished his book yet (no indicator of the quality, just my lack of time), but I wanted to start discussing some of the views he puts forward and inspires.

Socrates’ daemon (daimonion/daimon) was a kind of oracle that offered insight but not advice. [Interesting that many bots are Oracles: Robert Kendall’s Soothcircuit; Nicholas S. Roy’s iGod; The Pythia @ Winged Sandals; The Glide Oracle; Ron Ingram’s BuddhaBot…know more?] Socrates believed, apparently, that the daemon was an ‘independent supernatural creature’, which he described in The Symposium as being Love, an ‘intermediate between the divine and the mortal’. Socrates didn’t conceive of a daemon as being good or bad, but during his trial this spirit he referred to became the latter.

The ‘word daemon’, Leonard ventures, ‘however spelled, uncovers a provocative and useful dualism. An intermediary with another world doesn’t have to be beneficient. Yet neither is it compelled to be nefarious. It can be both, flip-flopping between positive and negative states–depending on context or perception, on the vagaries of polities, or the whims of the fickle masses.’ The next daemon to be summoned was nineteenth-century physicist James Clerk Maxwell’s Demon. Maxwell’s Demon is a hypothetical invoked to explore the second law of thermodynamics: the entropy of the universe is always increasing. Fundamentally, something has to be given in order to get (eg: energy to get heat). Maxwell’s Demon is a “… being whose faculties are so sharpened that he can follow every molecule in its course, such a being, whose attributes are as essentially finite as our own, would be able to do what is impossible to us.” (source). The Demon is basically a guard or valve that monitors molecules in two rooms (there could be an interesting juxtaposition with the Turing Test here). The demon opens the sliding door or valve to let the molecules move between them at appropriate times, thus heating one room and cooling the other without expending energy. [This is just my attempt at an explanation of a complex theory.] ‘The worst one could say’, Leonard continues, ‘is that the tiny demon represented wishful thinking on a gigantic scale: Wouldn’t it be nice to have little helpers doing our bidding, in willing defiance of the laws of space and time? Wouldn’t it be grand to extend our limited human abilities into new, otherworldy dimensions?’. (Sound like a good definition of a bot!)

Then in 1958 Oliver Selfridge proposed a program called Pandemonium. Daniel Crevier (an AI historian) apparently wrote that the program was like “Milton’s captial of Hell: a screaming chorus of demons, all yelling their wishes to a master decision-making demon”. Selfridge himself is credited as introducing the notion of ‘demons’ into computer science. But this is all background, the next demon was actual.

In 1963 Fernando Corbato and his team running the IBM 7094 had big success with their time-sharing systems, but now wanted to have processes that didn’t require ‘constant human supervision’. They created one that ‘endlessly scanned the 7094’s databanks looking for files that had been modified since its last scan’ (19). In homage to Maxwell’s Demon, Corbato wanted to name the autonomous helper Demon. Michael Bailey, a member of the team, apparently recommended the invoking of Socrates’ daemon. Leonard deems Corbato’s Daemon the ‘ur-bot, the primeval form to which all present and future bots owe ancestry’. He acknowledges that many programmers would scoff at this claim, but says Corbato’s daemon influenced generations of MIT grads who then created gamebots & chatbots.

Then Leonard cites Joseph Weizenbaum’s 1966 bot Eliza as the “Bot erectus - the first software program to impersonate a human being successfully” (33). By impersonation he refers to the chat capabilities of Eliza. Saying Eliza was the “first chatterbot”. Then Weizenbaum’s colleague, Kenneth Colby, is discussed in light of Parry (1972). Eliza was modelled on a Rogerian model of therapy while Parry was a paranoid schizophrenic.

The next step forward was with Michael Mauldin’s bot Julia (1989?), created for the TinyMUD created by James Aspnes in 1989. Julia was ‘a giant step forward for botkind’ (41), not only for her conversational ability and because she also performs services, she ‘inspired countless hackers to put together their own bots and unleash them in MUDs and MOOS’ (42). In the 90s then, Leonard continues, we have Kenneth Schweller’s bot MrChat, which resides in the MOO he created: College Town. Schweller’s code has been adopted by many botmasters creating bots for MUDs & MOOs.

I personally haven’t included daemons in histories of bots I’ve given in class. This is because I haven’t known about the others. I usually discuss software agent programming in general — explaining ’simple reflex agents’ (see Russell & Norvig)– but not the range of agents available (time constraints a large factor). I will include these agents from now on, but there are other pivotal points in bot history. What of Richard Wallace’s Alicebot (1995) for instance? If we are to describe bot history according to impact then Alicebot would be right up there. The software, which is highly accessible, has sporned hundreds of botmasters. It has also placed in the Loebner Prize four times (what kind of success this is an indicator of I’m not sure).

Indeed, the history of AI in ‘interactive drama’ hasn’t been touched on, for instance: Carnegie Mellon University’s Oz Project and more recently Mateas & Stern’s Facade (2005); or even generative texts: William Chamberlain and Thomas Etter’s Ractor (1984). We could also look at bots, indeed AI, through the perspective of how knowledge is represented, as Jorn Barger does in his Timeline of Knowledge-Representation. Heck, if we look at bots from a medial and arts type perspective, IF is a sibling. The text-based fiction with simple-reflex responses is highly relevant. So then we’d touch on Terry Winograd???s SHRDLU (1968-1970) (source: Montfort); Will Crowther and Don Woods’ Adventure (1975-6), Tim Anderson, Marc Blank, Bruce Daniels and Dave Lebling’s Zork (1978-1979) and so on.

And what about non screen-oriented bots, robots? They are part of the human journey to create autonomous beings, the research that goes into human-robot interaction is applicable to human-agent interaction, and alot of the agent software is and will be utilised in them. The history of robots — beginning with autonoma — is vitally important to bots (chatbots): Jaquet-Droz’s writing automaton (which apprently wrote ‘I do not think…do I therefore not exist?’ (source: Wood); Vaucanson’s duck and flute player (1739); Edison’s Talking Doll (1890); and the many Eastern autonomata — karakuri…and more recently the developments with robots: Rodney Brooks‘ space and doll robots; Cynthia Breazeal’s Cog and Kismet

Bots do of course fall under the mantle of ergodic literature, literary machines and so on. Discussions on the heritage of such works have always cited invocational game-like generators like the I Ching. Given the daemon/Oracle background of bots, it seems justified then to include the history of divinatory arts. Indeed, those studying divinatory arts could cite bots as another form.

It seems, to me, that there are the following approaches to the history of bots. One that views or conceives of the history of bots in the context of [list edited after initial pub.]:

  • efforts to understand ourselves;
  • efforts to contact an ‘other’ (fill in your own concept of ‘other’ here);
  • software & programming;
  • impact;
  • creating other beings;
  • creating human beings;
  • creative and literary developments;
  • simple-reflex agents;
  • literary simple-reflex agents;
  • Artificial Intelligence;
  • literary Artificial Intelligence;
  • Embodied Agents;
  • text-based fiction;
  • and of course, any mix of the above.

As Andrew Leonard acknowledges in his book:

The bot family tree is a confused and contradictory plant, a warped and twisted structure as unlike Darwin’s great Tree of Life as a blackberry bush is unlike a weeping willow. (23)

I like the idea, though, of bundling the antecedents in clusters, to acknowledge the range of forces that have impacted, indeed, been essential to the creation of bots. Bots would not of come about, nor will develop, without the influence of all those approaches. Whatever your view of bots, they are without doubt a confluence of many aspects of this world.

About Andrew Leonard
It was difficult to find some info about Leonard. He used to be a writer for Wired Magazine, the publishers of the book, and a senior technology correspondent for Salon. Leonard writes an article about bots in Wired, Bots are Hot!, pre the launch of his book.

About the Book
There are some reviews of the book: one by Holly Gunn; another one at NYT; ECRB; one by Rod Pollock and a short review by Bernard Cohen @ trAce.

Over to You
I haven’t covered the entire history of bots, agents, robots and so on. Just given an overview of Andrew Leonard’s agent paradigm. Indeed, I haven’t spoken about a definition of bots, which is often implicit in the choice of heritage. The first chapter of Leonard’s book is online at the New York Times (free registration req). If you get the book and are keen to start discussing it too, feel free to register on this blog and go ahead and post! But for now, I’d like to know your views on the history of bots, and I’d especially like to know more about significant gamebots.

21 Responses to “Bots, Demons & Dolls”

  1. 1 Mark Marino

    This is an incredible post. Let me reply a bit at a time. Here’s another Oracle Bot of a different kind:

    Go Here to Diana Slattery’s Glide.

    Also: regarding the history. Here are two other links that might help:

    Here’s an old timeline I had put together. I had Julia at 1990 for some reason, so it’s good to know there’s an earlier date.

  2. 2 Christy Dena

    Hey Mark,

    I do have Glide in, but I didn’t know the designer, so thankyou.

    Re: Julia. I have 1989 with a question mark because I’m not sure. Andrew Leonard doesn’t give a date (just the date of Aspnes TinyMUD) and neither does Janet H. Murray in Hamlet on the Holodeck; and neither does Leonard Foner; and neither does Maudlin in any of his papers or websites (that I’ve read)! So don’t quote me. I guess I should email Maudlin…

    Your history is great. It makes sense that SHRDLU is under bot history. I just read about it via IF and so presumed contextual ownership! And you’ve got Alicebot as 94. I just checked the AIML Overview page at and it says 95, do you have another citation?

  3. 3 Christy Dena

    And it also seems that Emily Short’s Galatea (2000) is a point where the history of IF and bots undeniably crosses. Although we can discuss the differences in a whole other post. Consider, for instance, this paper by Ruth Nestvold.

  4. 4 Mark Marino

    1995 looks correct, too, Christy.

    1994 applies to Barry DeFacto, Jackie, and the work of Thom Whalen.

  5. 5 Andrew Stern

    I met with Leonard in Berkeley in 1998 after his book came out, which discussed a project I worked on, Catz, for a paragraph or two. He was pretty into the topic at that time, but since Wired became a bit tired, I haven’t seen many more articles from him. The book never got much attention in academic or industry circles for some reason.

  6. 6 Christy Dena

    Oh yes! Catz, Dogz & Babyz (1992-1999)! See and These have been discussed in:

    · Hamlet on the Holodeck: The Future of Narrative in Cyberspace, by Janet Murray, 1997.
    · The New Media Reader, eds. Noah Wardrip-Fruin and Nick Montfort, 2003.
    · Information Arts: Intersections of Art, Science, and Technology, by Stephen Wilson, 2002.
    · Interactive Storytelling: Techniques for 21st Century Fiction, Andrew Glassner, AK Peters, 2004.
    · Bots: The Origin Of A New Species, by Andrew Leonard, 1997.
    · Avatars!, by Bruce Damer, 1997.

    I think they are important, AT LEAST on the impact side. Indeed, what about AI in movies (the crowds in LOTR) and so on? These are also important agents, but not ones we can talk to. Your Catz, Dogz & Babyz definately come under the banner though.

    It is such a pity that his book wasn’t well received at the time. Maybe now the botmaster critical mass has been reached (just playing with ideas) his book can be revived.

  7. 7 Rich Wallace

    One CEO of a popular software company told me that ALICE and AIML bots have an 80% market share for all bot deploted, owing to the open source nature of the code. Such market studes are hard to come by however, because there are several steps in the evolution of a bot.

    1. A bot appears on a web site or chat room.
    2. Downloadable code may or may be available.
    3. The code may be free, or charge a high royalty.
    4. The “learing curver”.
    5. Few companies release exact figures on downloads or sales.
    6, Bot generally have a finite lifetime.

    So the “world pouplation of bots” is hard to determine, let alone what percentage are using AIML.

  8. 8 Rich Wallace

    3a. Not everyone who downloads a bot, installs. it.
    3b. Not everyone who installs it, customizes it.

  9. 9 Rich Wallace

    7. Perhaps another factor affecting the world bot population is the growth of bot communities, which make it easier and easier for novices to sign up and chat with a bot, usually a clone of ALICE. But the audience of most of these bots is quite limited, and the lifetime is short (expcet for afew.)

    Couting bots is like couting celebrities. First, how do you define one? Second, do you count dead ones? Third, do the have to be starring in a major motion picture or just a part-time extra? Do they have to work for the screen actor’s guild? What I have only one line in one movie, am I celebrity?

  10. 10 Christy Dena

    I like your points about bot usage (botmasters & users). Indeed, there are plenty that download Alice, but not all install & customise it. Bots are spreading across cyberspace, but the audiences for bots (people who keep coming back) is still a relatively small and elite group. That is one reason why we’re planning on a BotFiction Comp, to encourage the use of bots for fiction (narrating a story, embedded in a story, a story emerging from a conversation…). There are lots of IF comps, but no Bot Fiction comps. There are comps about bots in general, but they are more exploring the personality of bots. [If I’m wrong, let me know.] We’re eagerly awaiting the upcoming bot narration script you’re bringing out for Alice, before the comp goes ahead. :)

  11. 11 Christy Dena

    From discussion on the Alicebot listserv (republished with permission):

    If you’re discussing Greek “daemons,” it’d be helpful to point out (if I recall correctly) that the Greek word for knowledge was “demon.” In Latin it was “scientia,” from which we get “science.” As Carl Sagan put it in his excellent book “The Demon-Haunted World,” a jurisdictional dispute is exposed.

    I like the reference to the _I Ching_ and other “generators”; that’s an interesting way of looking at AI programs. Kris Schnee

    Oh, and there are some citations of when Alice was cited in histories at Alicebot History.

    As a postscript re my last post. I should clear up that we’re not about to launch the Bot Fiction Comp, it’ll still be a bit of time away.

  12. 12 Dennis G. Jerz

    Since Will Crowther’s original version of Adventure was closely related to the geography of a real cave, IF in its origins had perhaps more connection to information than it appears. The cartography he did for the Cave Research Foundation is a completely separate intellectual project from the game that he created after he stopped caving, but both the map and the game are efforts to express the knowledge he had gained about the Bedquilt area of Colossal Cave (part of Mammoth Cave National Park, in Kentucky). Don Woods and the creators of Zork continued the underground theme, but they drew on their imaginations rather than experience. (Montfort does a great job describing how engineering and MIT lore supports the Zork series.)

  13. 13 Christy Dena

    Hello Dennis. Thanks for adding this point on conceptual heritage. If I may explore you point: On a similar vein, AI software is modelled on what we know about the brain — not the slimy curves but the interpretive systems. Both IF and bots then, came from urges to explore ‘information’: the former on how to represent it and the later how it is processed. This is probably why they continued into slightly different trajectories.

  14. 14 Christy Dena

    On the term ‘chatterbot’:

    Andrew Leonard cited conversations in the TinyMUD about Michael Mauldin’s Julia as the first use of the term ‘chatterbot’. Mauldin then, in his papers, was apparently the first to bring it out further.

    And here is a paper on the history of anthropomorphism by Peter Morse (a bit high-brow = inaccessible at times!).

  15. 15 Andrew Leonard

    Just wanted to drop by and say how fascinated I am to see this topic and such an informed discussion of all things bottish. If anyone is interested, shortly after the publication of the book I became a full-time employee of Salon where I have been ever since.

  16. 16 Christy Dena

    Hello Andrew!

    Thanks for coming by and thanks for writing the book on bots. What are your views on any recent bots/community/research/whatever? I’m interested in how you think things have changed, your impression on what is happening now, and the future (again).

    Also: Searching for your name with Salon I found another review of your book at

  17. 17 Andrew Leonard

    Well, it’s kind of tragic — but not long after my book was published, I became quite distracted with chasing another story, the free software/open source movement, which had only a tangential relationship to the topic of “bots.” So I can’t say I have kept up with the field in any kind of rigorous way.

    However, I was intrigued by the recent rash of stories about online poker-playing bots — to me, everything represented by the challenges of autonomous poker-playing bots followed directly along the lines I was pursuing in “bots.” It could easily have been another chapter.

    I think that advances in artifical intelligence have continued to move very slowly, but all the same, there have been lots of incremental improvements in software that have made autonomous or semi-autonomous software programs ever more reliable and useful. I also think that what I called the “technodialectic” is still in full force. Technical solutions to problems posed by autonomous software (spambots etc) only result in more sophisticated challenges. And so on and so forth.

  18. 18 Mark Marino

    Welcome, Leonard, would you flesh out some of your thoughts about poker-playing bots and how it fit with your conceptualization of bots?

  19. 19 Jeremy Douglass

    In 1999, I was crazy about the idea of Catz - not personally a power user (although I tried it and thought it was a lot of fun), but I found something really compelling about the fact of the online communities trading and designing and sharing pictures etc. - Petz automatically inspired these webring cultures that were different than a lot of the portal-cultures dot-com companies were trying to foster then. There was something that felt authentically folk about them, like the spontaneous outpouring of enthusiasm when people show you vacation slides or wallet photos.

  20. 20 Jeremy Douglass

    I’ve also been following the poker-bots stories with some interest - I love these running fire fights between the human and expert systems. Some related issues:

    The spam wars, in particular the bringing of Bayesian analysis to bear on the problem and the corresponding rise of supermutated prose spam and random content spam.

    The CAPTCHA wars, both the morphed-image text recognition battles (including what is essentially counter-counter-image-recognition) and simpler related obfuscation practices from javascript and image-embedding to simply posting emails such as jdouglass AT removethispart DOT ucsb DOT edu.

    Arimaa - a very fun game, but more importantly a response to the history of chess computing, and an explicit attempt to optimize a deterministic system to play to the strengths of human vs machine intelligence.

  1. 1 WRT: Writer Response Theory » Blog Archive » What is a chatbot? er, chatterbot?

Leave a Reply

thesis writing service