Adventures du jour!
Yesterday’s highlight was going to a darkly delicious movie, Bones and All; today’s was mostly AI related.
Bones and All was an interesting movie. It’s horror, but it’s romance, but it’s a road-trip movie. It’s slow and it’s not at all funny. It’s definitely not for everyone. It doesn’t even feel like a movie for most of its length, and it was impossible for me to track the progress of the plot through the normal writerly cues.
But I wasn’t bored for even second, and at the end I was hit with a rush of emotion at how beautiful it was and how ingenuously everything had been slowly set up and put together. The intended audience might as well have been me personally. I went back to the person who recommended it and told her my reactions throughout the movie, and she said she’d had the same ones.
I’m gonna say that if you liked Joy Division, then you’ll like this. You don’t have to like Joy Division to like it, but it has something of the same flavor.
This morning’s yoga I mostly skipped on, in order to walk to the bus stop with Ray this morning. Worth it.
On the walk, I worked mostly on releasing the hips downward so they wouldn’t cramp up. As often happens when I’m trying to convince muscles not to cramp up, it was a stop-and-go operation, walking until something tried to tighten up, then stopping to let to calm down and release. The rest of today has been pain-free, except for a massive freaking headache that’s probably sinus-related.
After Ray got back from school today, I bugged her to make pizza dough with me before she started doing other things.–She’s made some before, but that was a long time ago and HER mind’s gone through quite a bit of unnerving events as well, so she didn’t really remember it. She keeps buying these expensive packages of naan that’s not all that good to make mini-pizzas with.
I grasp that instafood, or near instafood, is often a good thing. But it still annoys me to have to spend cheese-level money on four stupid pieces of bread that don’t even taste that good.
So we made pizza dough, slapped it out, and turned it into sauerkraut pizza. Which is its own special kind of freaking delicious. Pizza, marinara sauce, sauerkraut, mini bell pepper slices, spicy salami, cheddar cheese: it was almost like the equivalent of a green olive and Canadian bacon pizza. I quite enjoyed it. I made the sauerkraut: bagged coleslaw mix, thin jalapeno slices, salt, time. I highly recommend making sauerkraut with jalapenos; anything pickled with jalapenos in it gets a yummy buttery-spicy kind of taste to it.
Writing: struggling with next class assignment that’s due soon. MEH. I’m at that phase of things. C’est la guerre.
AI:
–Worked on my AI class. Every time I come up against one of the optional labs, I have a mini-freakout, where I go, “They’re going to make me do programming.” It’s weird but it’s also not weird. A lot of programming things assume you know more than you really do, and “beginner” projects aren’t. I can’t tell you how many times I’ve ragequit on trying to learn programming for anything because of poor materials that answer no questions and give no context. Like, what IS linear regression? Blah blah blah, an answer that doesn’t explain what regression is, just that it’s a straight line somewhere in the middle of a bunch of data points. [Twitch] But I have Google for that, right? I got it sorted out; these labs are actually not bad. I just expect to have to face anger every time I start one.
–Talked to one of the AI guys about AI/marketing stuff. The last couple of weeks have been HUGE, with both a huge increase in public awareness and a huge increase in AI capabilities. We talked about stuff going on behind the scenes with his startup, plus stuff about ethics, not on the level of “should AI use other people’s work too train” but “how can AI be more transparent and accountable, to mitigate risks and set the public’s minds at ease?”
I have a different opinion than he does, mostly because of the way my mind works: I never feel like I understand something until I can break it or identify how to take unfair advantage of it. (“How could I break into this place?” is one the mental games I play when I go places.)
So when he says stuff about making the results of his AI products more transparent, I smile and nod, but I also go, “And how does this prevent assholes from doing exactly what you just said they would do?”–As a rule of thumb, it takes an order of magnitude to unfuck something than it does to fuck it up. (For example, racist hiring practices.) Making something transparent can tell you *what* has been deliberately fucked up, but it can’t stop it from getting fucked up. And I’d rather see more people focusing on that. But at least we were on the same page with how possible and likely it was that people were going to start doing shit to fuck up AIs. I hope I planted a bug in his ear.
–Played with the new ChatGPT AI from Open AI, a text-based AI that’s supposed to be one of the most sophisticated ones out there. The one from the other AI company I talk to already blows it out of the water. The same stuff that the other AI struggled with that initially looked like nonsense but turned out to be pretty cogent, took down ChatGPT multiple times (that is, asking it to give the meaning of “The Circular Ruins,” a short story about AI written by Jorge Luis Borges). The other prompt I like to use to test, asking for the system to tell me a fairy tale about subject X, then again with subject Y, to see if it can tell me two different fairy tales, was also a failure. (I got a lame fairy tale that was virtually identical except for the names related to X and Y.) I’m not working with any AIs related to music or sound, so I can’t speak to those, but image AIs seem far more sophisticated than text ones. Odd. I would have expected exactly the reverse.
–Read a bunch of articles/posts about AI today. What I mostly picked up today was that it’s probably going to be *very* hard to convince the courts that training AIs on publicly available data is against the law, and that it’s probably going to be *very* hard to make sure that laws aren’t written for the sole advantage of big IP holders like Disney, and greatly to the disadvantage of smaller-scale, individual artists.
–Of course Neal Stephenson pointed this out years ago in his SF novel Diamond Age, but if you think this is bad, wait until you can make this stuff 3d and hook it up to a 3d printer. Why buy stuff from a company, if you can just print it? WHAT will we marketers do? (O flow my tears, the serial entrepreneurs said.)
At any rate, it’s a conundrum. I intend to use whatever leverage and knowledge I get to argue for what I see as a better future, but I also get that it’s not going to be all hearts and roses. A writer friend who does both art and AI keeps arguing that this level of AI just illustrates that we need a universal basic income. He’s not wrong.