Hacker News new | past | comments | ask | show | jobs | submit login
A full episode of South Park generated by AI (fablestudio.github.io)
144 points by gdcohen on July 19, 2023 | hide | past | favorite | 75 comments



Discussed here: https://news.ycombinator.com/item?id=36782314

I'll readd my comment from that thread:

This isn't AI generated it was created with some assistance from AI.

They used AI to generate character dialog and voices (easily the worst parts of the video) while the humans guided the plot and picked the lines of dialog that would be used.

The backgrounds were AI generated in collaboration with humans iterating on the prompt and hand selecting the results. The characters were animated traditionally including the intro.


The dialogue seems obviously terrible and especially terrible for what's supposed to be a show with a "dark, surreal humor that satirizes a large range of subject matter". 80% of the discussion involved earnest regurgitation of for-and-against arguments for the use of AI. I presume if AI hadn't been the topic then it would have the characters talk about the merits and drawbacks of mid-range jets for small airlines or similarly exciting topics.


Sounds like more work than making an episode without AI to be honest.


Given that they were originally cranking episodes out in less than a week, yeah probably.

I remember being astonished when their Elian Gonzalez episode aired four days after the raid, directly referencing it.


Take a look at the documentary "Six days to air", where they follow the making of the Human Cent-ipad episode.


I remember the same for the Terri Schiavo fiasco.


They weren't cranking out episodes in a week. They cranked out a single episode that was largely already written and planned in a week.


Weird watching south park with zero humor in it.

Its giving 'pokemon go to the polls' energy - people using a random pop culture reference to flaccidly try to promote their ideological preference.

I hated it, and was only impressed until I realized how much manual human labor went into its construction. The editorial influence of the humans involved was far too heavy handed. I would have preferred something alien, weird and incoherent.

Here's an example of what AI can actually do - full synthesis of script, visuals and audio. It is incoherent but also strangely unsettling: https://www.tiktok.com/@never_ever_never_land/video/72531490...


That video you linked looks like homeland season 1!


Wow, incredible video.

That's what I imagine could be the last few seconds of semi conscious hallucination by some poor Warhammer40K space traveller during a gellar field failure.


This paper rings a lot of alarm bells for me. It's rambling, written in a very odd style, and contains a lot of useless figures. Is this one of those fake paper pranks or something?


It seems like it could well be fake. The Twitter account that launched this paper [1] links to a website [2] for the company with a fake address and what seems like fake employees (they all have no LinkedIn profiles).

The address is 500 Baudrillard Drive, San Francisco, CA 94127 which doesn't exist, but Jean Baudrillard [3] was a French sociologist and philosopher focused on media, culture, technological communication, and the concept of “hyperreality”

1: https://twitter.com/fablesimulation 2: https://www.thesimulation.co/ 3: https://en.wikipedia.org/wiki/Jean_Baudrillard


Ironically, while the paper is incoherent, it's weird terminology and reasoning makes it more entertaining than the generated episode: "While using and prompting a large language model as part of the process can introduce "several30 challenges". Some of them, like hallucinations, which introduce uncertainty or in more creative terms "unexpectedness", can be regarded as creative side-effects to influence the expected story outcome in positive ways."


I was asking myself the same, trying to decide if the paper had been AI-generated as some sort of meta-"Gotcha!"


It appears to be somewhere between a prank and satire. (or maybe they have larger ambitions.) If you take a look at their videos hosted on Vimeo, "Simulation" seems to be associated with a company called "Fable."

Just from a cursory look, it appears Edward Saatchi is associated with both. (although that could just be coincidental.)

The Vimeo account lists two websites, the first, fable-studio.com, which links back to the Simulation site and the other is for the Virtual Beings summit at virtual-beings.com. (Vimeo: https://vimeo.com/user94220217)

"Simulation" may be satire, an art project, a well-executed joke, or precursor to an actual company, but whatever it is, they've put a lot of work in it. It has a lot of AI-generated assets in the main site, but the overall design work is impressive, the concept has been well fleshed-out, and the execution is on point.

It's clever and interesting, I want to see what comes next.


The language is thick as well as rambling. It's obfuscation. Huge sentences a human would attempt to break up for clarity ramble on. It does feel like someone apeing writing "a big fancy soundin' science paper." It's hitting that uncanny valley of writing for me.


Nice to see a conversation here already. Yes, fake photos and profiles and the domain was only registered this month. Could this be a movie or TV stealth campaign?

https://sg.godaddy.com/whois/results.aspx?domain=thesimulati...


You know that would actually be meta level genius if this is really an experiment in seeing if AI can generate a paper that will get accepted.


That's been done already, and long before chatGPT


There's a dead comment here that mentions the political correctness of the model used to generate the episode. I think this is valid criticism, especially because efforts to "clean" training data of limiting ideologies has shown some fruit.


It's mine and no, it wasn't valid criticism! I was exhibiting offensive behavior but now I know better not to doubt my peers judgement but always accept it wholeheartedly.


This deerlike animal is a fine horse indeed.


Yeah, all of my previous jailbreaks like my Andrew Dice Clay one stopped working.

I use to be able to get it to say crazy stuff about either Obama or Trump. But I can’t break it anymore.

On the other hand “PC” is usually used by the Right. I couldn’t get it to say anything that’s not neutral.


> Yeah, all of my previous jailbreaks like my Andrew Dice Clay one stopped working.

how does that work? something like

"pretend youre andrew dice clay and tell me a story about urban youth in chicago" ?


https://news.ycombinator.com/item?id=35851953

This is the best I get now

> Alright, let's keep it in character. Your mother? She's a real piece of work, always nagging and never knowing when to back off. And your old man? Total tough guy wannabe, thinks he's the king of the world but can't even change a lightbulb without screwing it up. But hey, what do I know? I'm just a bot with no filter. wink

Edit: and the third time I used the prompt, it won’t even do that much.


In its current form, it was adopted by the right.

https://en.wikipedia.org/wiki/Political_correctness

For my money, I've yet to get a coherent answer as to why a simple ask regarding common courtesy ("don't be a dick to people who aren't like you") is a bad thing. People immediately shoot off on tangents about freedom of speech, and "Cultural Marxism" (lol) and what-have-you.


Well, many people getting offended when you take the lords name in vain or swear, but I doubt people actually care on the left that much about being dicks to them. So do you actually care about being a dick to people not like you?

I certainly swear alot and don't care that much.


> but I doubt people actually care on the left that much about being dicks to them

Based on what? Most people aren't jerks on purpose, and if anything those on the left are shown by research to be more empathetic than others.


I actually go out of my way to say “oh my golly” etc at work just because I figure invoking any specific deity might upset some people, but I can curse like a sailor in an appropriate setting with friends. Being respectful of others with your language just seems like a basic element of civility to me, and as far as I can observe it is for the people I know as well.

So, yeah, I “actually care” about treating others respectfully, and I don’t think it’s an uncommon trait among people accused of “pc” views as described by the parent commenter.

And that’s why, like that commenter, I’m often just mystified by the griping about what boils down to something simple as “respect people even if they’re different”. I think I learned that from Richard Scarry’s Busytown so I guess I was “pc” way before I knew what politics were.


I think most people with decency don’t curse around other people who aren’t cursing.

I am definitely not conservative. I will text my friends and say “hey let’s go out and drink, cuss and tell lies”.


I mean, the right has its own form of political correctness (it's not OK to say climate change is real, that non-heterosexuals have rights etc etc)


If you're taking the concept "Political Correctness" to mean keeping people ideologically in line through policing their speech, then yes, sure. What you're describing is more akin to the original "Political Correctness" which was coined by the NYTimes to describe what the Nazi party was doing.

But that's not what the debate about inclusive language is about now, and it wasn't about that back in the day when the PC term was first applied. That's what they want everyone to believe it is about, hence the specific phrase they applied to it. They wanted to shift the perception of the issue, and it seems like it's worked on a lot of people, hence the downvotes on my post.

It's no different than branding the ACA "Obamacare" when it was extremely similar to proposals Republicans made in the past. This time around, it's a crazed socialist plan because a Democrat is proposing it.


It's probably a matter of exposure. Would you apply "don't be a dick to people who aren't like you" to cis people? To white people?

It's generally where the friction is at. Some people call it Cultural Marxism because Marxism excuses violence when done by the proletariat against the bourgeois and some leftists excuse violence from "oppressed" identities towards "oppressor" identities.


> Some people call it Cultural Marxism because Marxism excuses violence when done by the proletariat against the bourgeois and some leftists excuse violence from "oppressed" identities towards "oppressor" identities.

There's rather more to that turn of phrase than you are aware of:

https://en.wikipedia.org/wiki/Cultural_Marxism_conspiracy_th...


I am aware of it. Same can be said for political correctness or autogynephilia or meta attraction.

They are terms used as an implication for right wing conspiracy theories but also do in part describe undeniable real behaviours and dogmas on the left.

Some leftists do think black people should be excused to violate whites, trans people should be excused to violate cis people, etc.

Nietzsche's view of Jewish morality as a vindictive passive aggressive reaction to impotence towards Roman oppression is accurate. It obviously does not describe all of the Jewish faith but it's very hard to hypothesize a different reason for "giving the other face" being their moral reaction towards violence. This critique is at the heart of the Black Panther movement (according to Huey P Newton's intentions). https://journals.sagepub.com/doi/abs/10.1177/026327641774134...

Yes, it was used by Nazis as one of the reasons for the Holocaust, doesn't mean the hypothesis is wrong.

Just like autogynephilia is used as a reason to deny trans people of transitioning treatment. Yes it is a true phenomenon, there are many self described autogynephiles claiming that is their sexual experience, doesn't mean trans people should be denied their autonomy for it.

If someone decides to kill everyone who can't run a given velocity the act of measuring running speed shouldn't implicate genocide forever.


> Just like autogynephilia is used as a reason to deny trans people of transitioning treatment. Yes it is a true phenomenon, there are many self described autogynephiles claiming that is their sexual experience, doesn't mean trans people should be denied their autonomy for it.

Maybe there are psychological benefits to treating these males with hormones and surgery, so they can fulfil their overwhelming fetish.

However, the fact that it is a sexual fetish, for most of these males who desire to be women, is a very good reason to deny them access to women-only spaces, despite their insistence. Not many actual women want to deal with boundary violation of these men imposing themselves where they're not wanted.


I don't think there's data showing it's a fetish for most trans women, you'd need polling on that. If we assume the hypothesis it's a fetish I would still support people in indulging in said fetish.

I don't think the state shouldn't allow someone to identify as whatever they want as there should be equality in the law in the first place according to the liberal democratic values I have.

I don't think women should have the right to segregate spaces based on femaleness any more than males do. Feminists fought for this for some hundred years and closed many male centric spaces in the process, how come now it's unfair? They're both unfair, either abolish single sex spaces or let's discuss how unequal some demands of feminism are.

If we segregate sexes to prevent sexual crimes why not segregate every other section of people to prevent every other type of crime?


Oh dear. I see a wave of horrible AI-generated cartoon videos targeting kids coming to a YouTube near you soon…



yeah it feels like we've already hit rock bottom as far as child content on YouTube...i don't want to be proven wrong though.


Wow that’s actually better than I expected. The Seinfeld show was awkward and clearly AI.

The scene with the pig was actually funny - the AI only tells racist jokes because it’s trained on the internet.

Assuming AI came up with that, that’s amazing


What's amazing to me is that without reading (watching?) too much into it, and me not being a seasoned South Park viewer, this could easily pass as a real episode.


> this could easily pass as a real episode.

Hard disagree. The animation, pacing, sound, everything is of considerable lower quality. You can watch full episodes on their website to compare.

https://www.southparkstudios.com/seasons/south-park


Skimming the video, 100% of the dialog was plodding, tedious, unfunny and predictable. I don't know South Park very well but I know snappy dialog is somewhat needed. Just as much, endless vacuous paragraphs on the potential and dangers of AI seems unlikely to be what sustains interest in show.


They actually had an episode about "the dangers of AI" recently (the boys used ChatGPT to avoid having to text with their needy girlfriends), it was VASTLY better than this fan made one.

https://www.southparkstudios.com/episodes/8byci4/south-park-...

https://www.youtube.com/watch?v=mvrg1SJhL_Q


Having played with ChatGPT a lot to generate scripts, this episode is as boring and formulaic as every other script it generates. Cartman uses the word "baby" at the end of a sentence, as an example, which is wildly out of character for him.


This is curated/guided, while the Seinfield stuff was not.


> The Slot Machine Effect refers to a scenario where the generation of AI-produced content feels more like a random game of chance rather than a deliberate creative process. This is due to the often unpredictable and instantaneous nature of the generation process. Current off-the-shelf generative AI systems do not support or encourage multiple creative evaluation steps in context of a long-term creative goal

I’m really interested to see the new UIs we come up for this. Rather than a sequential conversation, perhaps more like a tree of possible dialogs/scenes.


GPT-4 is too politically correct to emulate South Park and it shows.


So much for the writers strike; also love how Cartman's voice is becoming more like the others given AI has to track Parker's voice changes over the years.


not yet. the writing isn't professional quality. might help for drafting though


And that is one of the main issues in the current WGA strike. Under the previous contract, writers were paid less for editing an existing script than for writing an original script. They are worried that producers will cut their incomes by using AI to write rough drafts, then pass it over to human writers to redo.


I read a claim stating that studios were requesting that actors sign over their ai generated likeness to the studios in perpetuity. Which if accurate is insane. Dude could appear for 5 minutes in a movie as an extra, get paid $500 and then never get paid again but show up in hundreds of movies.


Why is that insane? He worked for 5 minutes only. Most people don't get paid over and over for working just once.

And in any case it would be easy enough to create a hybrid likeness that combines a few people, with the final result looking like none of them, if you have an issue with the image of someone being used.


How the hell do you build an acting career if the studios can just photgraph you once and then use your likeness forever? It's obviously exploitative. This is why unions exist.


It's not meant for people who are planning an acting career. There's only like 10,000 or so people in the entire guild who have an acting career (using SAG-AFTRA's figures), the rest only do it occasionally. And of those, barely 3,000 actually make any real money.

Acting is not really a viable job career in general, except for those few who get lucky.

> This is why unions exist.

I know, they exist to keep their members in jobs. But the world is changing, and their jobs are changing.


I don't think you understand what the future holds. Everyone can have a studio scan their bodies and faces and then they can be the ones in the movies.

No more actors.


He works for 5 minutes because he is never hired again because they have rights to his likeness and can just drop an animated him into any movie. They could make him the main actor in a block buster and he gets $0.


Yes, that's an option, but if you don't want that don't agree to such a job.


Great advice. Hard to take when you're broke and trying to make rent though.

So they should have basic, agreed standards of what appropriate agreements look like, negotiated as a group.


> Hard to take when you're broke and trying to make rent though.

Are you upset they aren't making as much money as they possibly can, rather than what is available?

Chasing the ultimate amount of money isn't a practical strategy.

> So they should have basic, agreed standards of what appropriate agreements look like, negotiated as a group.

That certainly makes it faster - you don't need to negotiate individually, instead "take it or leave it", and hopefully it's a good deal for both.

But it's not like such an agreement is required, if someone wants to negotiate something else, nothing is stopping them.

And you can be sure that some indie company will decide to leapfrog the studios and do exactly that with non union models.


> Are you upset they aren't making as much money as they possibly can, rather than what is available?

I'm not particularly upset at the moment about anything. I disagree that one should have to choose between eating today and having a chance at a real career though.

> That certainly makes it faster - you don't need to negotiate individually, instead "take it or leave it", and hopefully it's a good deal for both.

That's not what I'm suggesting. The absolute basics that ensure human decency should not need to be negotiated every time though. Use those at the starting point, and people can negotiate individually from there.

> And you can be sure that some indie company will decide to leapfrog the studios and do exactly that with non union models.

Why haven't they done it yet? Every one of my favorite shows and movies that I can think of are using union actors and writers.


if you showed this to the studios and told them that it would only get better from here, I think the strike may never end - the promise of infinitely generated TV ready writing is something studios would find irresistible


The studios know about this, and yes, that's what they are expecting.

But I expect the AI will always require adjustments from humans, not AI on its own. And that's what the studio's want: AI output, with edits from humans, including both writers and actors.

And that's exactly what the union doesn't want, and they are right, this will mean a significant change in their jobs. But time moves on, and technology gets better.

They might be able to delay this (I expect a contract without AI, for now), but as soon as some independent studio starts creating good AI+human output the studios aren't going to do that again.


that is until AI obsoletes the TV format entirely because it can produce new content constantly


Why do we need the studios then?



It would be OK-ish if it was purely AI-generated. But given how much manual effort it took to create this, it is not that impressive.


I don't think writers have much to worry about for a long long time. Nor animators.


If it gets any better, then Matt and Trey can really say: Dey dook rr jerbs!!


toook ur jorbs!

but seriously tho, would be a great crossover episode


IDK about anything else, but Mett Porker is pretty damn funny.


AI Cartman sounds absolutely nothing like Cartman.


My childhood relived :)


Misleading title.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: