48
votes
As its moderators remain on strike, Stack Overflow introduces "Overflow AI"
Link information
This data is scraped automatically and may be incorrect.
- Title
- Announcing OverflowAI
- Authors
- Prashanth Chandrasekar, Eira May, Joy Liuzzo
- Published
- Jul 27 2023
- Word count
- 1314 words
Previously on Tildes:
Thanks for the context. Looks like reddit isn't the only web forum with a dramatic saga playing out.
(late) update: Moderation strike: Result of negotiations
Before catastrophizing, I'd note the actual announcements on the page
The question is whether AI is going to totally replace all labor or just make workers more productive. Factories and industrial farming didn't replace everyone; they just made us more productive and massively boosted living standards.
If a writer can iterate off an AI rough draft, they may actually produce a greater quantity of better art! Anyone that has worked enough with the current generative AI systems knows they still need a human somewhere for oversight.
I think something important to mention is that while productivity and living standards can increase together, they don't have to. There's been a lot of literature on why exactly the rise in living standards after WW2 in the US occurred and the productivity increases were just one part of the puzzle. In comparison, take the past twenty years where productivity increases are ballooning just as much as they did 50 years ago, but boosts to living standards and real income have comparatively lagged.
But will the economic incentives align to actually promote writers who use AI to make better art? Or will they align to encourage mass-produced mediocrity which by virtue of their speed of creation crowd everything else out? Look at today's TV landscape. Good shows with good writing are comparatively rare and junk is common, but junk can and does make money so that's why there's still so much of it. So even just relying on pure human effort, our batting average of making "great art" is pretty low.
I'm incredibly skeptical AI will actually improve things on this front -- outside the hobbyists who do not rely on their creative work paying the bills. The only "hope" I see is, thus far, US courts have refused to grant copyright protection to AI produced images and writing.
I'm also really glad that works created solely by AI can't be protected by copyright in the US. There's definitely structural concerns about how the incentives will shake out, but I personally feel there's a lot of promise.
YouTube creates endless amounts of unwatchable junk, but high-quality content creators have replaced many traditional sources of information and media around the world. I'd like to think that there's an increase in content creation across the board. Sure there might be more AI junk, but there will hopefully be some AI gems too.
This AI "anime" project from a VFX team shows some of the promise I think. Their process was very different and much more complicated than hand drawn or CGI animation, but the result is artistically notable I think.
https://youtu.be/_9LX9HSQkWo
I think one criticism of Corridor crew here, is that if you're making content for commercial purposes (as Corridor crew here are) then you're profiting off the work of other artists (in this case the folks who made Vampire Hunter D Bloodlust) without fairly compensating them. It would not be possible to make the animation Corridor made if Vampire Hunter D Bloodlust wasn't around. To quote Niko:
I don't quite agree. Rather than being inspired by old anime, this type of technology actively requires screenshots from other IP to function. I would have been much more on Corridor's side if they hired an artist to develop a style for them (even a style inspired by other works) and use those to train the AI rather than blatantly taking screenshots.
Edit: Here's a pretty levelheaded (mostly positive) take on the tech from a real animator: https://youtu.be/xm7BwEsdVbQ?si=OS1BuCb8RPqC_rvL
It was my understanding they released the video for free, they charging for lessons on how to achieve what they did with the tools currently available. Is my impression incorrect?
Even if they released the video for free, they're youtubers, so they almost certainly still profited from it. But honestly even if they got zero benefit, it's still immoral to use someone else's work without compensation like that imo. It's toeing the line when you use a big dataset full of multiple works, but it's much more obviously unethical using just one work imo.
You can choose not to have specific videos monetized on YouTube AFAIK. And I disagree, for example riffing off of other artists work is what makes our music community so vibrant and amazing. So long as you are respectful to the original artist and aren't attempting to profit from it without consent.
I would argue that training a machine learning model on one speciric show falls squarely into NOT this category. I'm all for artists riffing off each others' work, but I don't think taking someone else's work and feeding it into AI is the same thing at all.
Whether it's their chord progression into a new instrument, or their art into some other generative tool, I have to agree to disagree with you.
This is an issue in the current writer's guild strike. Studios would like to have AI generate the rough draft, then hire a writer to fix it, because writers get a lower rate for revisions than for creating the thing in the first place. It ends up not actually saving any work, because it's a lot easier to make something right in the first place than it is to fix what the AI put out.
Seems they should pay the same rate regardless of what the employee is wasting their precious time with. It's crazy to me that revisions pay less...
I think the scariest (and most likely) outcome is: in the near future, AI doesn't completely eliminate many labor roles, but it does increase productivity---which essentially means that many labor positions are eliminated, with no replacement in sight. Labor continues to get squeezed, but remains too atomized to agitate for radical change. (And even if we did achieve, say, basic income, what would we do with our government-sponsored leisure time? All the creative disciplines will be automated. What's even the point anymore?)
I'd argue that if we had leisure, people would create art for the sake of art. I recently watched an interview of a pair of sisters in their 70s who'd been knitting lace by hand all their lives. The interviewer asked one of them what she did with the beautiful lace pieces she made once she was done with them, and essentially she had tens of thousands of hours of hand-knitted lace items in boxes under the bed. She said that if someone wanted to give her money for them, sure, but she didn't really worry about it. She hoped that one day when her kids were cleaning out her house there'd be enough lace for all of them who wanted any.
So even now when there's a need for money to pay the bills people make things for the joy of making things. If/when we reach luxury automated space capitalism, people will still be making art because it has meaning to them and because they want to express something to the world.
I think this is a misconception of how and why people create art. People are naturally artistic, yes, but I think that the artistic impulse is, to a large extent, bound together with the social impulse. Most people do not (consciously) create art except in some kind of social context; that is, most people intend for their art to be appreciated by other people, even if only by family or close friends. Without that, I think most art disappears. A heavily atomized society like the one we live in does not provide conditions for the artistic impulse to flourish. And it only gets worse if the zone gets flooded by AI-generated art. Human attention is limited, and the more attention that gets paid to AI art, the less will get paid to human art.
There are also feedback effects; e.g., AI art will reduce demand for artists and art teaching, which means that artistic knowledge will be lost over time. We have precedent for this sort of thing, so it is one of the consequences of AI art that I am quite certain will take place.
They made society on aggregate more production. Plenty of people died in the meat grinders that were 18th/19th century cities. So not every (or even most) individuals at the time benefited.
To be clear, the strike is by entirely unpaid volunteers. It makes it a little different, since the only thing causing them to strike is their strong opinion that corporate's actions are wrong, not anything else.
As much fun as I have chatting with AI, I think there's no substitute for a real flesh-and-blood human being at least at this point in the tech's development. I mean, why am I here on this forum talking to you about it when I could be talking to chatGPT instead? Take for instance online multiplayer in video games. Sure, you could spend your time shooting bots, but there's no substitute for the sheer variety of outcomes when fighting human players versus the "perfect play" a bot might achieve. Also, on the user's end, victory against a bot is quite hollow when compared to victory against another person.
Consumers of entertainment are not going to be moral arbitrators as AI tech infiltrates into our creative industries, so, I expect the industry to go through changes similar to how automation changed assembling cars, but I do hope there remains a "craving" for human-led or made products. I don't know how you engender this kind of cultural zeitgeist, however.