Haven’t been following it too closely but this specific demand struck me as doomed to failure. (MBA here stands for “minimum basic agreement” as I understand it)
regulate use of artificial intelligence on MBA-covered projects: *AI can’t write or rewrite literary material; can’t be used as source material; and MBA-covered material can’t be used to train AI.
That’s never going to fly. If AI has utility in the creative process it’ll be used, if it is worse than a human, it won’t. Either way it is a problem that solves itself.
Of course the WGA wants to try and protect its members, everyone understands that but it strikes (ha!) this outsider as about as likely to succeed as sailmakers striking over steam propulsion or sealing wax makers striking over end-to-end encryption.
Would I be bothered if something I watched was a product of AI? Can’t say that I would.
Judging by some of the dross that I’ve watched recently either AI is already at work and screwing it up (in which case the market will give its verdict) or human writers are badly in need of a kick up the arse in terms of doing better, I’m not sure AI would be worse in some cases.
Anyone else with strong feelings either way regarding AI in the creative industries?
As I’ve said elsewhere, our copyright laws were written in a world where riffing off someone else’s work took considerable labor of one’s own. If I want to write like Kurt Vonnegut, I’ve got to spend many hours reading his books and practicing. Whereas if I just want to plagiarize him, it’s far easier for me to do so.
The law reflects that difference: I’m allowed to riff, but I’m not allowed to plagiarize. Riffing involves enough of my own labor that I can legitimately treat it as my own work; plagiarism doesn’t. Riffing is allowed; plagiarism isn’t.
AI changes that calculation. Riffing with AI is trivially easy, involving very little labor.
Our copyright laws absolutely need to catch up with this. I hope the strike raises awareness of this issue and slows down the process long enough for the legal protections to catch up.
Sure, that labour saving is absolutely on point and we can all think of myriad examples of similar historical advances. Where AI is different is the outsourcing of creative thought beyond the human. We just haven’t been here before.
But to your other point. let’s say I ask AI to write a novel about X in the style of author Y.
Assuming you don’t try to pass it off under their name or try to fool anyone, I’m not sure where the harm lies. Nor what legal protections could ever be put in place to stop anyone doing it.
The technical contractual thing here, as I understand it from speaking with friends in the industry, is that AI will be used to generate initial drafts, and then human writers will be hired to rewrite that material. AI generated writing is shitty, so the writers will need to rework it heavily to make it acceptable.
But contractually, the original writer and the rewriter are different roles, and are paid differently. What happens if there is no original writer, and only rewriters? Basically the studios get the rewriters to overhaul the AI generated material and produce an initial functional script at the reduced rewrite rate.
That, to me, seems entirely reasonable as something worth objecting to. It’s not about whether AI can or cannot barf out something superficially resembling a script, it’s about the studios using AI to eliminate an entire category of writing and get the same writing services for less money.
Long enough to jam sabots in the gears of progress…
Sure! That! As always, a cogent and on-point reply!
The harm is that wealth will–as it has in so many past advances–move from the laborer to the owner. A writer’s work now becomes grist for the mill, and it’ll be very difficult for individual writers to compete with a factory of AI-enhanced rewriters, concentrating wealth even further than it’s already concentrated, and removing one more autonomous profession from the world.
Obviously, a lot of people think the industrial revolution was a goddamned lark and a picnic, and see no reason for anything to go any differently this time around. I disagree: while the changes may ultimately be beneficial, we should take the time to ensure safeguards, minimizing the harm to individuals and society of these changes. Changes to copyright law could be part of this.
As for possible changes? Here’s one: an AI may only use material for training purposes when there’s an express agreement with the material’s owner.
in that situation, seeing as they are doing a rewrite job why would a rewrite rate be inappropriate?
That’s pretty much the story of technical progress. If AI can do that successfully why would it not be used?
Because the AI script is unusable shit and they’re using it to get the writer to generate an actual first script while paying the rewrite rate.
Sounds reasonable in practice but I see very little chance that such a rule could be policed successfully.
This got discussed quite a bit in this thread (which I admit I checked out of while the discussion was still going on):
Note that the writers apparently have two concerns about AI:
- “studios … using AI to generate new scripts from writers’ previous work”
- writers being “asked to rewrite draft scripts created by AI”
The latter doesn’t seem inherently wrong to me. But if the writers want to try to use their collective bargaining power to exclude certain duties from their job description, I guess they can try?
The former has a little more weight IMHO. I’ve heard of visual artists complaining that art-generating AI has been “trained” on their work and produces results that essentially rips off their style and elements of their work. And it seems to me that they have a legitimate beef, and that writers’ fear that something like this could happen to them is likewise legitimate.
It would be a de facto killing of generative AI. You are just not going to get a depth and breadth of output skills without vast amounts of training data.
If the AI script is unusable shit then it is a self-correcting problem. No-one will use it for that purpose.
So how do you limit access to the vast amounts of genius level training data that already exists in the public domain? No-one can get the OK from Dickens or any of the Bronte’s
ChatGTP (not sure which version) used 300 billion words of training data. That’s equivalent to around three million novels.
As a routine part of her constituency work our local member of Federal Parliament sends out a newsletter of run of the mill standard and quality.
Recently there has been a significant improvement in the literary merit of the communications. Now this could be because she now has a larger office staff and an improved media manager or they could be farming the task out to ChatGPT.
I’m not sure this fazes me because it’s hardly creative work. If it is her message just polished and dressed up and consequently is easier to understand then fair game methinks. All manner of pamphlets, notes, corporate work and short form literary grist might be similarly improved.
But the notion of generating some substantive work via AI and claiming it as original does rather disquiet me. I recently did a course on travel writing. Now there’s a profession which will soon and maybe deservedly go the way of flint knapping courtesy of ChatGPT. Conversely I also did a course on profile writing and if that literary form is taken over by bots we’ll be the worse off for it.
I’d be all for stipulating that it was AI generated. That wouldn’t bother me.
Now I’d be the other way around completely. I’d think the profile would be very well suited to AI and travel writing less so.
Who really knows? One thing is for absolute certain though, AI generated work is out there now and we are unwittingly consuming it.
It is not being used for the purpose of creating a filmable script, it is being used for the purpose of getting the human writer to generate a filmable script under the fiction that the work is a rewrite and is therefore payable at a lower rate.
A comment I saw recently on Facebook (from someone replying to someone talking about how bad AI generated art is) was something like “You have seen bad AI art. You have also seen good AI art, you just don’t know it”.
But it is a rewrite is it not? Whether generated by AI or not.