California bill (AB 412) would effectively ban open-source generative AI

[Bohandas]

I was answering to @Bohandas in speciffic who claimed to easily be able to afford an artBreeder and a NovelAi subscription. Paying for both is about $50 to $60 a month, enough to pay a few art commissions.

Novel AI is $10 per month and Artbreeder is $7.49 per month, which totaled together is less than the price of one commission, which from what I've seen seem to start at around $20 for a single image. And there's no world where I'm paying $20 for a single image even if AI goes away. I could get multiple books or computer games for that $20

 

[Bohandas]

Why are you all assuming people are actually using paid online AI service? There is strong evidence they are a minority (albeit a large one) use in image generation compared to models one runs on his own computer for the price of electricity.

I'd love to do this but I don't have the technical expertise nor a good enough computer to run anything more advanced than GPT-2.

I can run GPT-2 locally thanks to some scripts that one of the guys on GitP helped me put together,, but even that just chugs along and I have to run the smallest version of the model otherwise my computer crashes as often as not.

 

[Bohandas]

That's a different discussion, and a bit of a whataboutism. Copyright laws aren't perfect but at least they protect small creators from being crushed by big media conglomerates.

My experience is the reverse. I've seen multiple small time artists get persecuted for drawing art of things like pokemon. Cultural touchstones like that really ought to be open to everyone, but alas our shared culture is owned - in a legally binding sense - by large corporations. People should be allowed to own their own culture.

 

[MoonSong]

I have to second this @MoonSong . That is in fact the literal point of training models. Training a model is the process of feeding information into the model so that future queries relying on the model interpret those queries to include that new information.

I understand the point you are trying to make. But the answer a query to an LLM is not that different in most ways to a query put to a human being in that regard. Both the LLM and the human have a 'perception' (to use your term) and give a response based on their experiences and what they know about the world.

AI doesn't have experiences, nor needs or feelings. All of these color the way we "read" media and content. There's a book I love. The first time I read it was at seven or eight. I later did a book report on it during High school. Despite it being the same literal book, seven year old me read a way more fantastic and colorful book than older me. Sixteen year old me knew more about colonial times and naming conventions so that book was way more mundane and less fantastical. Of course I prefer the version my child version read. In a similar way, I can no longer listen to American Idiot and The Black Parade without crying my eyes out, these two hit me in a different way than five or six years ago, because up to that point I loved these albums as a source of joy and inspiration for cool adventure hooks and stories. Now, these are a source of catharsis when I feel like missing my Dad. That's the kind of "learning" and perception that AI is incapable of. It cannot create significance nor reinterpret new meanings into older experiences.

Legend tells that when Gary Kasparov lost to Deep Blue, he went into depression, and later got better when playing with his kids. He also went for his favorite food and had a date with his wife, only to realize that the machine couldn't do any of that. It was just a piece of silicon unable to feel or care.

 

[chuckdee]

AI doesn't have experiences, nor needs or feelings.

The point is, you're only right about two of these. It does have experiences and these color the way that subsequent data is interpreted and contextualized, just as it does for us.

 

[Clint_L]

I'm experimenting with testing the parameters of ChatGPT and NotebookLM in various ways - bear with me; we're all doing it in my field of work, and just had another pro-D session on it all day yesterday. Anyway, I asked Chat+ to comb through my Internet history, using my Google login (this one), and try to fashion a review of this trailer that approximated my style.

I have my own thoughts, but I'm interested if anyone who is familiar with my style/tone from thousands of posts on this forum would think it was me:

Against all odds, the new Superman trailer doesn’t just look decent—it looks like it might finally understand the assignment.

James Gunn seems to be doing something we haven’t seen in a while: presenting Superman as an actual person, not a monument. There’s color. There’s warmth. There’s Clark Kent walking with people instead of hovering above them in a beam of moody light. It’s not ironic. It’s not edgy. It’s… sincere. And for Superman, that’s a swing worth taking.

David Corenswet has the right look—clean-cut without feeling plastic—and he moves with the kind of presence that suggests he’s in on the emotional core of the character. It helps that the trailer isn’t afraid of tone: we get a bit of retro optimism, a flash of cosmic weirdness, and what feels like a world that still has civilians in it. I caught myself thinking, this actually looks like a place Superman would care about.

Yes, it’s busy—Metamorpho, Mr. Terrific, Guy Gardner, Hawkgirl—but Gunn has earned some trust when it comes to ensemble work. If anything, I suspect the supporting cast will make Clark feel more grounded, not less. And let’s be honest: if DC wants to rebuild their universe without reboot fatigue, anchoring it around a version of Superman people actually like is the smartest place to start.

I’m not ready to declare victory. This could still go sideways in post-production, or get lost trying to please both the Reeve nostalgists and the Snyder holdouts. But for the first time in a long while, the idea of Superman doesn’t feel apologetic.

And that, in itself, feels like progress.

 

[Jfdlsjfd]

Not convinced, TBH. More AI mannerisms than yours.

 

[Bunch]

The point is, you're only right about two of these. It does have experiences and these color the way that subsequent data is interpreted and contextualized, just as it does for us.

It also has energy needs. I'm sure someone has threatened it in instances and seen if/how it reacts.

 

[Clint_L]

I think it has some elements of my style, like the way I tend to structure forum posts journalistically, with fairly short paragraphs intersperced with one sentence paragraphs to sum up a key point. I probably overuse italics; this piece certainly does, but not in the same way I would. I don't think the tone is that close to mine - I couldn't imagine writing the sentence "I’m not ready to declare victory," for example.

I do basically agree with the content, even if a lot of it is kind of generic and addresses aspects of the trailer that I don't particularly care much about.

I would give it a "C+" as an impersonation. But given what it had to work with...not bad. I don't think my job is in danger yet, though.

 

[The Firebird]

I'm experimenting with testing the parameters of ChatGPT and NotebookLM in various ways - bear with me; we're all doing it in my field of work, and just had another pro-D session on it all day yesterday. Anyway, I asked Chat+ to comb through my Internet history, using my Google login (this one), and try to fashion a review of this trailer that approximated my style.

I have my own thoughts, but I'm interested if anyone who is familiar with my style/tone from thousands of posts on this forum would think it was me:

Against all odds, the new Superman trailer doesn’t just look decent—it looks like it might finally understand the assignment.

James Gunn seems to be doing something we haven’t seen in a while: presenting Superman as an actual person, not a monument. There’s color. There’s warmth. There’s Clark Kent walking with people instead of hovering above them in a beam of moody light. It’s not ironic. It’s not edgy. It’s… sincere. And for Superman, that’s a swing worth taking.

David Corenswet has the right look—clean-cut without feeling plastic—and he moves with the kind of presence that suggests he’s in on the emotional core of the character. It helps that the trailer isn’t afraid of tone: we get a bit of retro optimism, a flash of cosmic weirdness, and what feels like a world that still has civilians in it. I caught myself thinking, this actually looks like a place Superman would care about.

Yes, it’s busy—Metamorpho, Mr. Terrific, Guy Gardner, Hawkgirl—but Gunn has earned some trust when it comes to ensemble work. If anything, I suspect the supporting cast will make Clark feel more grounded, not less. And let’s be honest: if DC wants to rebuild their universe without reboot fatigue, anchoring it around a version of Superman people actually like is the smartest place to start.

I’m not ready to declare victory. This could still go sideways in post-production, or get lost trying to please both the Reeve nostalgists and the Snyder holdouts. But for the first time in a long while, the idea of Superman doesn’t feel apologetic.

And that, in itself, feels like progress.

Not that familiar with your style but I use the LLMs a lot and I can clock this as generated text pretty rapdily. Hard to describe why--its not the em dashes or the use of italics alone, but those contribute. The diction generally is very reminiscent of the responses I get though.

Note--I've not done any blind tests. But I see this stuff on webpages relatively often and feel in many cases I can tell.

I suppose it doesn't feel like someone has reflected on, edited, and chose how to present this content in a concise way. Wordy in the wrong places, lacking the right specific details. It's simultaneously clean and vague.