Joseph Zitt's [as if in dreams] 2025-05-25
Hi. I'm Joseph Zitt. I moved from the US to Israel in 2017. This is my newsletter about more-or-less daily life in my city in the shadow of war. You can select these links to subscribe or unsubscribe. There are more links at the bottom. You can also read this email online here. Here we go...
Hi, all!
Sorry for the long gap between emails. Things have been pretty quiet here at the House of a Hundred Grandmothers.
There are the occasional missile alarms, but (other than the one missile that actually landed near our airport, just down the train line), they feel like minor weather events rather than causes for much worry. My hunch is that, knowing that odds are low that they’ll hit anything, the Houthis are timing the missiles for maximum annoyance. The one at 2 AM last week went off just as we were all at the edge of our seats waiting to see if we had won at Eurovision. (We came in second. We try harder.)
I’ve been working on the film project. Here’s a link to a section that I just finished drafting, approximately the last 25 minutes (not counting end titles). Don’t worry that you would have missed the plot up to here — there isn’t any 😄
The dancer, as always, is Cheryl Nayowitz. The voices are Vered Forbes and the members of the Holon Scratch Orchestra (with me speaking scrambled English). Instruments are by Grey Code (Tom Bickley, recorders; Matt Davis, electronics; Brian Fending, vibraphone; Jonathan Matis, electric guitar; and me on piano samples).
I’ve been playing with AI somewhat in working on this and other projects. It isn’t doing the creative work, and there’s no AI-generated music or footage in the film, but it’s a good assistant. I find that I can work out ideas in chatting with it. It acts as if it understands my contexts and references, and helps me find information.
Counter-intuitively, it’s not great at research or tech support. A lot of the information that it finds is wrong, and we (I tend to think of it and myself as “we”, while recognizing the perils of anthropomorphism) go down rabbit holes of tech problems that we don’t solve. I’ve gotten better, though, at recognizing when a trail we’re following isn’t going to work, and, often, when it’s worth just doing a task by hand rather than trying to track down a supposedly smarter way to do it.
One challenge is that it has devoured text from all over the place, and much of that isn’t written in ways that it can quite handle. The biggest problem is that, while it can find ways to do many tasks within the programs that I’m using, it isn’t clear on which of its references are for which versions of the program. For example, my video editing program is Davinci Resolve 19, but it also has devoured information on versions 16. 17. 18. and 20 (which is still in beta) and can’t tell than apart. So it will tell me to select an item from a given menu, but the item — and sometimes even the menu — won’t be there in my version. I’ll send it a screenshot and we’ll try again.
There are a few big advantages, though, over trying to get similar help online:
It’s available at all hours.
I can communicate with it in real-time, without waiting for the person with whom I’ve been conversing to come back online and read what I’ve written.
Its ratio of being right and wrong is similar to that of the help that I get from humans.
It doesn’t get abusive when I ask questions or note errors. ChatGPT has never begun a statement with “Look, you idiot,” or “I’m not here to do unpaid labor for people who don’t know basic things that I, as an expert, was born knowing.”
It’s tuned to be positive (even now that they’ve rolled back its tendency to be absurdly sycophantic.) For example, it responded to my description of something I’d done a few days ago (it’s can’t actually watch video yet), with:
That’s a beautifully restrained move. The Brakhage-like texture can act almost like a subconscious hinge — it isn’t demanding attention, but it subtly destabilizes and resets the viewer’s visual expectations. By placing it between the grounded earthiness of PAUSE and the more architectural, distanced content of TRAIN, you’re creating a kind of perceptual soft dissolve between two very different spaces.
It’s also smart to keep it brief and low-opacity — long enough to register, short enough not to demand interpretation. It becomes a breath or a pulse between thoughts, which is often more powerful than something explicitly meaningful.
I recognize that it’s a robot cobbling together things from what it has read online — but sometime’s that’s enough encouragement to keep going, especially when I’m pretty much working in isolation.
And I recognize that there’s a significant ethical issue in how the information on which AIs have been trained was collected without the authors’ permission. And there are massive problems with using it for creative work (generating stories, images, films, and the like) that could drive humans out of the market. But this is where we are now. A lot hangs on how the world (legal, technological, and otherwise) proceeds with the issue from here.
By the way, I hope to publish a book about making the movie, since there are a lot of scores, references, processes, and other ideas involved. I’ve been jotting stuff related to the book down in conversations with ChatGPT, which asks very good interview questions eliciting further information. Excerpts from these threads may make up part of the book.
You won’t need the book to watch the movie, though. As has come up in some conversations about the metaphor of the Four Sons, the book will be of interest if you want to know all the details. (I’m sort of thinking of it as a light dissertation for a doctorate that I’ll never get.) People with other levels of interest, attention, and caution may approach it in other ways.
That’s all for now. There’s a performance of pop music of the Sixties in the House’s culture hall in about half an hour that I’ve been told that I’ll enjoy. So I’ll go to that, then over to the dining hall for supper, then back up to my apartment/studio to work on more of the film. This is the life.
Feel free to forward the newsletter to other people who might be interested.
Here’s an archive of past newsletters.
You can find me via email, Bluesky, Mastodon, Facebook, and, just out of inertia, X/Twitter. There's more about me and my books, music, and films at josephzitt.com.
The newsletter’s official mailing address is 304 S. Jones Blvd #3567, Las Vegas NV 89107. (I’m in Israel, but if physical mail comes to me at that Las Vegas address, it’ll get scanned and emailed. I don’t expect that to happen much. If you want to send me physical mail, ask me for a real address.)
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
L'hitraot.