I can think of two things off the top of my head: 1) an assistant to deal with my computing needs, and 2) a system to examine literary texts, movies, and other expressive texts to determine, A) whether or not they exhibit ring-form composition, and if not that, then B) what form do they have.
Computing assistant
I’ve already written a post about this: What do I want from my AI Assistant? [control, that's what]. Here’s a chunk of that post:
The fact is, I’m wedded to my computer and to the internet, email and world-wide web. I really couldn’t function very well without them, not as an intellectual. And for the most part I don’t have to spend all that much time fiddling around with things in order to keep them working. But I do have to spend some time. And, yes, I probably could use some changes. But I don’t have the skills I’d need to make those changes, much less the time.
It's obvious that I need an AI Assistant to take care of all of this. Some years ago I sketched out ideas for a PowerPoint Assistant I could control through natural language. I also imagined that what I was thinking about for PowerPoint could be generalized:
The PowerPoint Assistant is only an illustrative example of what will be possible with the new technology. One way to generalize from this example is simply to think of creating such assistants for each of the programs in Microsoft’s Office suite. From that we can then generalize to the full range of end-user application software. Each program is its own universe and each of these universes can be supplied with an easily extensible natural language assistant. Moving in a different direction, one can generalize from application software to operating systems and net browsers.
Back then – the notes originally date from 2002-2003 – the technology we’d need to do that didn’t exist. Now it does.
Who’s going to control these AI Assistants? The end-users or the MegaCorps?
I don’t have much more to say about that at this point beyond observing that I think privacy and security will be real problems here. If I were to think about writing detailed specs for that I might start with this working paper I prepared some years ago about how my use of personal computers has changed over the years: Personal Observations on Entering an Age of Computing Machines.
Determining the formal structure of texts
The thing is, this isn’t a deep problem, this isn’t rocket science. There are these discussions that talk about AIs that will one day find a cure for cancer, figure how to make a practical fusion reactor, discover a grand unified theory, and solve climate change. Determining the form of a literary text isn’t like that. It’s not easy, but it doesn’t take anything like genius either.
I specify ring-form composition in particular because it is something fairly specific to look for. I think that’s easier than simply requestion: Tell me the form of this text. I’ve spent a lot of time looking for ring-form composition, in narratives, poems, and movies and blogged about it quite a bit. I’ve prepared a number of working papers on it as well.
What makes it tricky is that I can’t come up with a list of specific indicators that can be quickly identified as signs that the text has a ring-form. Nor can a specify exactly what feature to look for. It varies from one text to another. The only general thing I can say is that the text have this general form:
A, B, C...X...C’, B’, A’
The first section of the text is echoed by the last, the second is echoed by the next to last, and so forth, and there is a central section that serves as a turning point.
I have a post where I presented ChatGPT with a text of “St. George and the Dragon,” which does have ring-form, and asked it to analyze the text. The results of that experiment are, at best, inconclusive. The form is more obvious than in post texts, and the text is a short one.
What would it have done with Shakespeare’s Hamlet, which is known to exhibit ring-form? I have no idea. James Ryan has identified ring-composition in 26 of Shakespeare’s plays. I’d like an AI that could verify his work, or at least rough out a description which Shakespeare experts could then check.
And so forth and so on through every text in the canon, however you want to identify the canon. But, by all, non-canonical texts as well. I’ve found ring-composition in a manga by Osamu Tezuka, Metropolis (certainly in the Japanese pop-culture canon), and in Obama’s “Eulogy for Clementa Pinckney,” which is not normally within the compass of specifically literary texts. I’ve also found it in films, such as the 1954 Japanese film, Gojira, or the “Pastoral Symphony” episode of Disney’s Fantasia. Films present a particular challenge as LLMs can’t view them and I suspect we’ve got a way to go to create AIs that can view films and parse the themes and action.
Identifying ring-composition is one thing. Identifying other formal structures is something else. You want to identify formal structures that appear in text after text, whether verbal or filmic. Structural description may not be rocket science, but it’s not obvious either. You have to compare texts with one another to see what makes sense. “What makes sense,” that’s a vague methodological prescription if ever there was one. But it’s the best I can do in a short post.
I could go on and on. But this is quite enough to state the problem. Perhaps I’ll say some more later.
No comments:
Post a Comment