Blog For Free!
Archives
Home
2004 February
My Links
AIML.info
tBlog
My Profile
Send tMail
My tFriends
My Images
Sponsored
Blog
|
| The Popestress |
| 02.18.04 (3:08 pm) [edit] |
The Internet startup that hired me went by the name of [url=http://www.enjoyventure.de/dt...]'moomax'[/url] and promised to offer 'maximum personality'. This was to come in the form of so-called 'avatars' - artificially intelligent 'virtual personalities', for which they had all kinds of wild plans, in the form of image-ladden PowerPoint presentations. Those critters were supposed to operate e-commerce stores, stand in as company representatives on the web, even front dating services that worked through mobile phones. What's more, they were to be 'the virtual pop stars on the Internet'. And moomax, of course, was to be the market leader in that sector.
When I joined, they had about 25 people working for them. Most of them were supposed to do client aquisition and marketing, but they also had a guy called the 'knowledge engineer' (an ex-lawyer who had been on sabbatical before he took the job) in charge of the 'knowledge base' (or 'brain', as they liked to call it in the PPPs), a couple of journalists who were to feed said 'brain' with knowledge (and hopefully, through their contacts, generate some press hype on the side), three secretaries (one for each of the company's founders), an IT guy, a psychologist, and two screenwriters charged with character development. Since I happened to know one of the journalists, and he had brought up my name in a meeting, I ended up trying to design the personality of the first of the 'virtual pop stars'.
Turned out to be tough going, though. There was no idea whether this character should be male or female, Martian or Balinese. As the moomax founders went out hunting for venture capital, that had been considered to be a mere detail. Now that they had been successful, and the company's operations were in full swing, it was supposed to 'emerge' from brainstorming meetings and 'dynamic group processes'. So we met in small groups, and we met in large groups; some meetings were chaired by our 'knowledge engineer', and some by our 'psychologist'.
Those were the best. That guy did all kinds of 'psychological' stuff to get our 'creative juices' going: he let us cut, fold and twist index cards into intricate little sculptures. He burnt incense. Tapes of tribal drumming provided the musical background to our sessions. Countless hours were spent with collecting the most harebrained ideas anybody could throw up on a large whiteboard. At one point, consensus was almost reached that the first of the virtual pop stars was to be God Himself. All that in the name of 'artificial intelligence'.
Meanwhile, I found out some things about the 'artificial intelligence engine' they were going to use. I wasn't even supposed to bother, since my contribution was to be the 'character framework', and the actual technology was treated as some kind of a secret. But I'm the curious sort, always interested in novel computer programs, and I was quite well-versed in finding information on the web by then, even though I had never meddled. Soon enough, I got to download a trial version of that 'AI engine' to my own box, and started playing with it.
Do you know about chatbots? I had never seen one before, but these days they are somewhat more common, so chances are you've already met one while surfing the web. Chatbots, or 'bots' for short, are computer programs meant to simulate dialog partners. For reasons I didn't know about back then, they are not very good at that job. For reasons I also didn't know about, they are better at that job than any other kind of existing 'artificial intelligence' application tasked with the processing of conversational input in natural language.
What I [i]did[/i] know after the first trials was that such a bot could reply to sentences I typed and entered, but didn't make enough sense for the whole process to be called a 'conversation'. It would say something like:"I'm a giant electronic brain," and when I asked, "What is that?", it would ask back: "What do you mean by 'that'?" After I found some more examples on the web, it became clear to me that those 'artificial intelligences', on the whole, weren't behaving very intelligently. And we were supposed to make 'pop stars' out of those?!?
Still, there was something about those... things. I had never seen anything that came any closer to my old idea of 'interactive text'. Yes, it was flawed, but nevertheless, I was sure that I was looking at something that could be.... something. A brand new medium. The text that was 'interacting' with me was badly written - but maybe something could be done about that. What if there were some as-yet-unknown tricks, techniques, rules that, once discovered, would make those clumsy 'conversations' more life-like, more convincing? After all, when films were invented, they were - apart from their sheer novelty - very boring, static artifacts, because nobody had an idea of camera positioning and movement, cutting, montage, and other things that would once make them so much more dramatic and effective. It took a couple of decades of experimentation until those early 'moving pictures' began to resemble what we now know as 'movies'. What if those 'artificial intelligences' I was looking at represented a similar case? Anyway, if what I could see was all we had to work with, I was sure that we could forget about creating 'virtual pop stars'.
I brought up that topic in our next meeting. I said that if the technology we were going to use was of such limited capability, the least that we should be doing was to design our character around those limitations. Let's start with the real, and make bugs into features. Nobody was trying to hear that, as everybody was still outdoing each other with outlandish concepts. So I wrote up a memo, adressed to our 'knowledge engineer', and cc'd to the company founders. In essence, it said that instead of talking about building a Ferrari, we should be talking about building a Model T. That got no reaction, either, and more and more I felt that I was wasting my time. And I don't like wasting my time, even if somebody is paying me to do it.
The company's christmas party was the last straw. We were nearly 30 employees by then, lounging in a fancy restaurant, eating excellent food and drinking as much Chateau-Neuf-Du-Pape as we could swallow. The father of one of the secretaries made an appearance dressed as Santa Claus, tortured us with self-smithed verse praising moomax and its ingeniuos founders, and distributed presents. After the ordeal, we lit up fat Cohibas, leaned back and looked very 'new economy'. Note that this was December 2000, the Internet bubble had long since burst everywere else, we were trying to convince people to buy non-existent brains, and we hadn't even agreed on what our first 'artificial personality' should be like. It was too much fun for me to stand, so I went home early and wrote my resignation letter. In hindsight, I suspect that Santa made me do it.
A week after that, a decision was reached about what the first 'virtual personality' would be: 'The Female Pope'. Yeah, right. The Popestress. Maybe all that incense and tribal drumming really got to those people - I don't know, I wasn't there. Anyway, the idea was that they would create a real ruckus by promoting a novel Christian sect, The Federation, and then launching the Popestress on the Internet on Easter sunday, right after the real, male, pope in Rome got through his "Urbi Et Orbi" thing. Imagine the controversy, imagine the headlines! And they really did it. It went down like water off a duck's back, widely unknown as the lowest point in the history of artificial intelligence. After that, moomax burnt through the remaining venture capital, and officially went belly up in December, 2001.
Lucky bastards. They got out, while I got hooked.
|
|
|
| |
| Where I'm from |
| 02.18.04 (7:03 am) [edit] |
I've always produced (for) media. I did journalism, publishing, screenwriting, music production, some graphic design, some TV shows, even some advertising and PR work. However, producing text was what I started out with, and know best how to do. I assume that all media products can be reduced to an underlying text - whether it's manifested in the pages of a movie script, or in the concept of an ad campaign, or in the symbols of a musical score, or even in the little labelled blocks passing through the edit window of some A/V sequencer program, it's all text to me.
I've been using computers since the mid-80s, and as a user, I enjoyed staying on the cutting edge. However, using the then-current versions of Word, Pagemaker, or Cubase never got me interested in finding out how these programs actually did what they did. Or, for that matter, who did them. Although I always held the firm belief that those people refered to as 'programmers' where artists in their own right (yes, even those who gave me Word 3.0 for DOS), I couldn't care less about how they accomplished their feats, just as long as they kept coming up with the next, better generation of media production programs.
As time passed, however, I got quite frustrated with the limitations of linear text as a medium. I wanted the reader to be able to somehow interact with my writing, or better, I wanted my writing to react to the behavior or mood of the reader. I had the vague idea that this should somehow be possible using computers when, came 1992 , I read about Doug Lenat and the [url=http://cyc.com]Cyc[/url] project in an American magazine named MONDO 2000.
Cyc was described as an 'artificially intelligence', a computer program that would have 'common sense'. Dang, I thought, itīs already in the works! I completely misunderstood what they where trying to do - I thought they were developing a tool for me to write text that could somehow 'argue' with its reader. I was really excited about that prospect, and badly wanted for them to get that application finished, so that I could be one of the first writers to use it.
In the meantime, I discovered the Internet and hypertext. Although I readily appreciated its use as a powerful research tool, as a medium, it was't at all what I had been looking for. Sure, it was non-linear, and that was wonderful in its own way, but the price that I, as the author, would have to pay for this non-linearity seemed to be that I would have to completely give up the idea of providing a coherent 'narrative', of 'telling a story'.
Regardless of the media I had happened to use, telling stories had always been what it was all about to me. And all the stories I really like are recursive - they eventually lead me back to the point they started off at, and when I get there, it still is the same point, but [i]I am different[/i]. It's in this integration of 'recursion' and 'expansion' where the 'sweet spot' of successful storytelling lies. Conventional, linear storytelling emphasizes the 'recursive' aspect - the author can 'expand' only along one particular axis. Hyperlinked web pages, in contrast, can expand almost infinitely, but I could see no way to control recursion. That wasn't what I had in mind.
I ended up using the net as a big and increasingly important source of professionally relevant information, whithout ever trying to meddle in the 'online world'. This went on for years, and probably never would have changed - but then, one autumn day in Y2K, I got a phone call. Someone I didn't even know hired me to help in creating an 'artificial intelligence'.
|
|
|
| |
| What's in a name? |
| 02.17.04 (7:09 am) [edit] |
"Art & Intelligence - huh?" Well, if I put those together, I end up with the term 'artificial intelligence' - AI for short. Not many people seem to be looking at the 'AI problem' that way, but I do. And I'm not only looking at it from an armchair-based, 'philosophical' point of view, but also - I could even say: primarily - from an implementation point of view: My general idea is that, in order to create artificial intelligences, I should try to combine the mushy, intuitive, generally 'un-scientific' methods commonly associated with 'art' with the rigourous, formal, 'scientific' methods of 'intelligence' research. And that's what I'm about.
Some definitions seem to be in order. The 'artist' part warns that no definitions can be set in stone, because everything needs to stay flexible for the work to be able to grow, but the 'scientist' part argues, successfully, that programming computers is impossible without some definitions to start off with - after all, in order to make distinctions, a computer has to be told exactly what is '0', and what is '1'.
So then, I'll need a definition of 'art'. I'll google for this, using that fine new 'define:something' feature, and pick the one I like best as a starting point. Which happens to be: [url=www.svsu.edu/mfsm/educational/glos strm46.html]"Objects or ideas created by humans which tell/show what we are thinking or feeling." [/url]
As a definition of 'intelligence', I'll start with this one:
[url=www.cogsci.princeton.edu/cgi-bin/webwn]"The ability to comprehend; to understand and profit from experience." [/url]
From those two, I'll construct me a definition of 'artificial intelligence': "Objects created by humans that communicate thoughts and feelings, comprehension and experiences."
This definition immediately rules out two classes of objects commonly associated with AI: those that operate in your car or in your camera to help you with breaking on wet tarmac or with focusing your lens on your mother-in-law, and those that are meant to evolve to a point where they eventually become superior to mere "human" thoughts, feelings, comprehension, and experiences. It's not that I don't find those kinds of objects interesting, it's just that I don't aim at creating them.
What's in, then? To put it short: a new medium. Right - to me, an AI is a medium, comparable to a book, a movie, a record, a video feed, a magazine... you get the picture. Media are all about communicating human thoughts and feelings, comprehension and experiences, and from how I see it, an AI should be doing just that, only in a whole new way. Over time, I plan to elaborate on this idea in this blog, but for now, I'll let it just sink in. Should be enough to seperate mine from everybody else's approach to creating AIs, anyway.
|
|
|
| |
|
|
