2y ago

Fable Studios created an AI-generated episode of South Park... during a Hollywood strike over AI.

techcrunch.com Maybe showing off an AI-generated fake TV episode during a writers' strike is a bad idea | TechCrunch

It's a complex time to be in media — but it doesn't take much to see this is probably a bad time to show off an AI that makes whole TV shows.

So, how much money do you think Matt and Trey are going to sue them for?

You're viewing a single thread.

65 comments

Just watched it. Writers have nothing to worry about for now. I do admit I laughed once, though.
- I took a look and it's honestly a lot further along than I was expecting in terms of capability. In all honesty, for low level conent this is already surpassing the minimum necessary and I can already imagine greedy, low effort art thieves going all in on these and jaming out completely shows. And I expect people will watch them, or at least tolerate some of them.
  
  I'm rather impressed by how coherent it was. It had themes, distinct characters, a plot arc, and so forth. And some very nice meta humor. I don't know how much of this comes entirely out of the LLM scriptwriter and how much was prompted in, but even assuming that this was done from a human-created outline it's still a big step.
  
  Its so wooden and all the jokes replaced with generalized statements. Did you actually watch it? Most low grade Youtube content knocks this out of the park.
  The only thing these media companies will be doing by replacing a single writer with AI is making their content closer to the static noise floor of content that comes out of Youtube and similar sites already.
  
  I can imagine this being a problem for low effort youtube kid shows, like the ones that caused the whole elsagate controversy.
  
  AI generated YouTube channels are already a thing and bring in millions of views.
  
  Ugh I've left autoplay on other some science videos while indie other stuff and it took me a long while before I realized it had progressed to ai jargon space videos. So fucking annoying
- You are making the same mistake I see a lot of people make when it comes to AI, which is looking at the status quo as a snapshot rather than a change over time.
  The last widely reported on AI generated 'show' was the Seinfeld one from...checks notes...a few months ago.
  The leap between what that was a few months back and this here is quite something.
  So your "right now" may be true for today, but quite possibly by as early as the end of this year there will very much be something to worry about.
  (Though really, there still won't be much to worry about, as the future will almost certainly be AI plus human efforts, not either or.)
  
  I think you’re making the same mistake as people who thought self-driving cars would be here 5 years ago. You can’t just extrapolate out technological progress. The relatively easy things get solved first and relatively quickly but we may need a decade to solve some of the most challenging scenarios.
  
  Self driving cars were here five years ago, which is when Waymo first had driverless cars on roads. Tesla had a wide release of FSD 'Beta' three years ago.
  And there's a gulf of a difference on the speed at which hardware that has an 8 year average refresh cycle grows in a market and software that can reach a hundred million users in 3 months.
  
  You’re right. We all love our fully self-driving cars and by 2026, chatbots will write longform narratives so beautifully, we won’t even need cars because we’ll all be transported anywhere we want to go by the magic of books.
  
  Though really, there still won't be much to worry about, as the future will almost certainly be AI plus human efforts, not either or.
  Think the concern is AI+humans means a lot less humans needed to do the job
  
  There are a few fields where there's capped demand so extra supply would mean less humans.
  But I think people will be surprised by just how much of our economy is capped by supply, and what happens to niche demand as supply rapidly increases.
  The people most in trouble are the ones that really suck at what they do, and whose only job security is constrained supply.
  But at the same time, lowering transactional costs (in the sense of the essay "the nature of the firm") will mean a lot more opportunities for small and medium entrepreneurship around passion side gigs suddenly being economically viable as full time gigs.
  In reality, the groups most screwed long term here are going to be larger corporations who lose the advantages of scale but are still weighed down by the hindrance of slow moving bureaucracy.
- My thinking is that from a studio's perspective it may be like a proof of concept that AI can get close enough to do what they care about make a passable imitation that gets buts in seats that will generate ad revenue or ticket sales. Fundamentally they aren't really concerned about producing quality material as long as it sells, so if the AI can get them to something kind of good its likely worth their attention. I think that's what writers and actors are concerned about and that is why even an unfunny south park episode is a threat. Fable can say their work is research all day long but their goal can easily change the second a studio shows up with a check in hand.
  Also it is not clear here is how much human editing and tweaking was done after the AI was finished with it's part. I suspect people kind of helped the AI get to a final product, but without them disclosing their procedure it's hard to know.
- The Danny DeVito one got me the first time.

65 comments