June 3, 2003
Methinks I see some crooked mimic jeer
I caught up with Michael in Atlanta on my way back from ACH/ALLC. Michael and I got to talking about (among other things) programs to generate stories, poems, and other creative texts. He mentioned that he benefits from looking at thoughtful symbolic architectures for this sort of thing, such as Minstrel, but really finds nothing of interest in the statistical approaches taken by systems like Gnoetry and Ray Kurzweil’s Cybernetic Poet.(Correct me if I’m misrepresenting you, Michael!) On the other hand, I do think there’s something to the way that these statistical systems manage to knit together a new sort of voice – something that makes the process of the generator interesting to consider – and I’ve written about this a bit in an article that is forthcoming in The Cybertext Yearbook 2003. Interestingly, a paper by Greg Lessard at ACH/ALLC described a limerick generator he had developed, called VINCI, so this topic is still a current one.
Our discussion led to questions about how, in this specific case of creative text generation, symbolic/rule-based AI can be integrated with statistical AI to achieve some of the advantages of each. …
The curious thing here is that a symbolic approach is not just a technique, but also a framework, a way of thinking about how to approach a problem at a high level – and similarly for a statistical approach. It isn’t as if you can just naturally plug in a symbolic AI module to a statistical AI module and have a working hybrid poetry generator. You have to figure out how these are going to work together. Getting your prosody guy and your meter guy and your sonnet tradition guy to sit down and try to produce something might, likewise, be pretty difficult.
Unfortunately, I don’t have much but questions about this matter so far. Let’s take the specific case of generating poetry. One guess about how different “modules” (or “agents,” if we’re talking about a Minskyesque “society of poets”) might operate would involve the “high-level” agents (dealing with theme, say, or perhaps narrative in the case of a narrative poem) having control over activating what the “low-level” (prosody, meter) agents do. But it’s obvious to me, from my experience as a poet, that poems are seldom if ever written by people in this way – a word can turn the theme as easily as the theme the word. That’s not to say that a “society of poets” approach wouldn’t work, just that it doesn’t seem right to organize it in the way that might at first seem reasonable.
I don’t write poetry or story generators, and I’m not even very widely read on the topic, but both the problem domain and the computer science issues are of great interest to me. Perhaps Michael and some of our poetry-generator-generating readers, if there are such, can chime in about whether my architectural/framework question is a meaningful one, and, if so, whether there are good guesses (or URLs pointing to good guesses) about how to answer it.
June 4th, 2003 at 1:22 am
A side comment – your link to Kurzweil’s Cybernetic Poet led to me find his “cyber art” site, including a downloadable version of the infamous Aaron painting AI! Wow.
And, further surfing revealed that Naughty Dog, a game company, is using high-performance Lisp to make complex behaviors for agents in their games. This is super-interesting, will need to investigate.
June 5th, 2003 at 5:25 am
The GDC presentation from Naughty Dog this year included a brief discussion of their LISP usage, which if I remember correctly, amounted to “we liked it and we’re invested in it now so we’ll keep using it, but it has enough drawbacks we wouldn’t recommend anyone switch”, if I remember the presentation correctly.
Some details are at the end of this set of powerpoint slides. My general impression is that this is exactly like every other scripting language experience; there are advantages to coroutines and interpretation, but language differences really aren’t all that significant (as has been suggested by others in other threads of this blog).
June 7th, 2003 at 1:03 am
I’ve written a few poetry generators, but I always use a markov-chain approach to it, coming as I do from the side of poetry less formal. One such example is my Flash Poem Generator at http:www.lewislacook.com/pogen3.html, which again uses a randomized Markov chain based on user-input. Another such project was Fort! Da! at http://www.lewislacook.com/FortDa/, which searches internal texts for a string containing words or phrases input by the user. The internal texts it searches are organized in subset; each subset is chosen based on the user’s system time. A kind of slantwise poem generator; here the architecture is a bit skewed.
Yes, the architecture of such systems is compelling, isn’t it? It’s one of my favorite toys…
bliss
l
June 7th, 2003 at 1:04 am
as for AARON–I downloaded it a while back, and to be perfectly honest it was too figuritive and predictable for me…
bliss
l
June 10th, 2003 at 7:31 pm
In addition to discussing story generation architectures, I harangued Nick about why he’s doing a Ph.D. in computer science that’s disciplinary computer science (read, research in things that conservative computer scientists recognize as computer science) rather than interdisciplinary computer science that treats his interest in interactive fiction (IF) as a first class research endeavor. Having just finished a Ph.D. in CS where my research was AI-based art and entertainment, and working to establish (some) credibility for a practice that deeply combines AI research and art making (credibility gaps existed on both sides of the fence), I want to see Nick pushing a similar interdisciplinary agenda. Nick retorted that such interdisciplinary work is hard to set up (admittedly, my situation was unusual and lucky) and takes longer to complete – his way he can get the knowledge, practices, methods etc. of a CS Ph.D. relatively quickly and then begin applying it to his Real Work (poetry, IF). I retorted that new computer-based art forms require deep research done in the context of art making, not the application of business-as-usual CS. He said that his current research, while not perhaps directly applicable, is giving him deep skills he can use for doing IF-specific research later. And besides, he’s enjoying what he’s doing now independently of his work as a poet, IF author, and scholar. I have a strange compulsion where what I do must be One, not Many. Since the One and the Many have been battling it out at least since the pre-Socratic philosophers, we left it at that.
June 11th, 2003 at 12:11 am
I think my colleagues and professors in computer and information science do treat interactive fiction as a first-class reserach topic – but as a topic in another discipline. My interests are really never all going to fall within a single academic discipline, even if we manage to come up with a new one, as I hope I we (and others) will. I enjoy both translating Latin poems and designing and analyzing algorithms, for instance, and I’m not exactly unique in having interests along both of these lines. So I’m very pleased to have a context for study that allows me to pursue non-CIS interests and activities while structuring my CIS studies so that I can learn things I otherwise would not, and hopefully make some advances in the field as well.