5 ms·

Can’t you replace a lot of Hollywood with AI now? What am I missing?

by rayiner 8mo ago

onetokeoverthe 8mo ago
[dead]
CyberDildonics 8mo ago
Prove it
bandrami 8mo ago
Not if you want to copyright the output
- dctoedt 8mo ago
  > Not if you want to copyright the output That gets tricky: To the extent that an AI is just a tool — along the lines of a trained pair of hands executing a human prompter's specific, detailed instructions — the human prompter might qualify as an "author." From the U.S. Copyright Office in January 2025: "The Office affirms that existing principles of copyright law are flexible enough to apply to this new technology, as they have applied to technological innovations in the past. It concludes that the outputs of generative AI can be protected by copyright only where a human author has determined sufficient expressive elements. "This can include situations where a human-authored work is perceptible in an AI output, or a human makes creative arrangements or modifications of the output, but not the mere provision of prompts. "The Office confirms that the use of AI to assist in the process of creation or the inclusion of AI-generated material in a larger human-generated work does not bar copyrightability. "It also finds that the case has not been made for changes to existing law to provide additional protection for AI-generated outputs." https://www.copyright.gov/newsnet/2025/1060.html https://www.copyright.gov/newsnet/2025/1060.html (emphasis and extra paragraphing added).
  - rayiner 8mo ago
    Interesting, thank you for sharing. That’s surprising to me though when you see the reasoning it makes sense.
  - bandrami 8mo ago
    The closest analogy is to a music producer sampling public domain audio. The composition will be protected by copyright but each individual sample will not. Every single CGI rendered frame of Shrek is protected by copyright because it was human authored. It they used a diffuser to make Shrek 7, the individual frames would not be protected by copyright but their arrangement into a movie could be. That's a hugely different legal situation (for instance, if I chopped it up and made my own remix of it that would be protected).
    - bandrami 8mo ago
      Though this is complicated by the fact that the LLM initial training may have been massively illegal (or at least massively tortious). There's still a bunch of legal shoes to drop, one way or the other
ben_w 8mo ago
Not "a lot" of Hollywood, not yet. Short clips are doable, but the coherency duration before it goes weird just isn't there yet, unless you're OK with very short segments (like, single-digit seconds) between each AI-equivalent of a cut. That will probably improve, but the tech isn't there yet for "a lot", it's like CGI in the 90s rather than CGI as it is today. That's aside from how AI output isn't copyrightable, which may or may not matter depending on if your goal is money or propaganda.
- fragmede 8mo ago
  You haven't been seeing what's been coming out of models from China. Hollywood is the Wiley coyote that hasn't looked down yet. They're turbo fucked.
  - happymellon 8mo ago
    > You haven't been seeing what's been coming out of models from China. Citations? I haven't, and would be genuinely curious if it is better than the AI short clip slop that appears on YouTube and really just needs to be taken out back and shot. I have no doubt that its coming, I just haven't seen it yet.
    - ben_w 8mo ago
      I think there's two things contributing to YouTube slop, the earlier generations of free model being one, the other is more fundamental: most people using these tools for slop only care about advertising money, they neither know nor care about how much of a difference artistic taste makes (hence also why there's so many YouTube shorts which are just a clip from a random TV show with inappropriate music added, sometimes as a mic-drop, sometimes just obscuring the dialogue, but how even are YT shorts monetised?) FWIW, there are models which are better than the majority of the YouTube slop I've seen*; but even then, when I show short clips from e.g. Google's "Flow" generator to others, those others find the results unimpressive. Also, the voice range is weird, my experience playing with them it had a much harder time of creating a decent working-class British accent than Dick Van Dyke. However, amongst the various issues Flow has, the most obvious is length, because even though Flow lets me "extend" a clip, it's not got enough coherency between each (8 second?) segment when I do. The second biggest issue is that, as with Stable Diffusion before it, it quite often produces stuff you just don't want. Junk dealers don't care about that and use the first result, an artist can afford to do 10 attempts and pick the right one, and will care to, too. That's something I can at least expect will be solved, or has been on non-free tools (control nets etc.) even though I don't expect slop dealers to ever bother with them. If each 8 seconds of video was actually Hollywood quality (they're not, in any sense including resolution) and they were coherent when extended (likewise, they're not), as there's 750 times 8 seconds in a 100 minute film, it would still be an existential threat to Hollywood if those segments cost $1333, because that's a million dollars and in the cost range of "micro-budget indie", rather than the tens of millions that even normal-indie films cost (before marketing etc.) or hundreds of millions that blockbusters cost. I have no idea how video generators scale up in cost with time or resolution, let alone quality, so I don't want to make guesses on how much extra compute (or indeed training data) would be needed for a coherent 2 minute cut-equivalent at 4k. As I don't like the epistemic collapse we're already witnessing with clips from video games being passed off as coverage of warfare, I very much hope this is one of those things which fails to scale up. * although even with early models, if you wield them as an artist rather than as a slop merchant you get interesting output; you may remember the Harry Potter Balenciaga videos from a few years back, same channel and you can see how things have changed: https://www.youtube.com/@demonflyingfox/videos https://www.youtube.com/@demonflyingfox/videos

Can’t you replace a lot of Hollywood with AI now? What am I missing?

by rayiner 8mo ago

Can’t you replace a lot of Hollywood with AI now? What am I missing?

onetokeoverthe 8mo ago

[dead]

CyberDildonics 8mo ago

Prove it

bandrami 8mo ago

Not if you want to copyright the output

dctoedt 8mo ago
> Not if you want to copyright the output That gets tricky: To the extent that an AI is just a tool — along the lines of a trained pair of hands executing a human prompter's specific, detailed instructions — the human prompter might qualify as an "author." From the U.S. Copyright Office in January 2025: "The Office affirms that existing principles of copyright law are flexible enough to apply to this new technology, as they have applied to technological innovations in the past. It concludes that the outputs of generative AI can be protected by copyright only where a human author has determined sufficient expressive elements. "This can include situations where a human-authored work is perceptible in an AI output, or a human makes creative arrangements or modifications of the output, but not the mere provision of prompts. "The Office confirms that the use of AI to assist in the process of creation or the inclusion of AI-generated material in a larger human-generated work does not bar copyrightability. "It also finds that the case has not been made for changes to existing law to provide additional protection for AI-generated outputs." https://www.copyright.gov/newsnet/2025/1060.html https://www.copyright.gov/newsnet/2025/1060.html (emphasis and extra paragraphing added).
- rayiner 8mo ago
  Interesting, thank you for sharing. That’s surprising to me though when you see the reasoning it makes sense.
- bandrami 8mo ago
  The closest analogy is to a music producer sampling public domain audio. The composition will be protected by copyright but each individual sample will not. Every single CGI rendered frame of Shrek is protected by copyright because it was human authored. It they used a diffuser to make Shrek 7, the individual frames would not be protected by copyright but their arrangement into a movie could be. That's a hugely different legal situation (for instance, if I chopped it up and made my own remix of it that would be protected).
  - bandrami 8mo ago
    Though this is complicated by the fact that the LLM initial training may have been massively illegal (or at least massively tortious). There's still a bunch of legal shoes to drop, one way or the other

ben_w 8mo ago

Not "a lot" of Hollywood, not yet. Short clips are doable, but the coherency duration before it goes weird just isn't there yet, unless you're OK with very short segments (like, single-digit seconds) between each AI-equivalent of a cut. That will probably improve, but the tech isn't there yet for "a lot", it's like CGI in the 90s rather than CGI as it is today. That's aside from how AI output isn't copyrightable, which may or may not matter depending on if your goal is money or propaganda.

fragmede 8mo ago
You haven't been seeing what's been coming out of models from China. Hollywood is the Wiley coyote that hasn't looked down yet. They're turbo fucked.
- happymellon 8mo ago
  > You haven't been seeing what's been coming out of models from China. Citations? I haven't, and would be genuinely curious if it is better than the AI short clip slop that appears on YouTube and really just needs to be taken out back and shot. I have no doubt that its coming, I just haven't seen it yet.
  - ben_w 8mo ago
    I think there's two things contributing to YouTube slop, the earlier generations of free model being one, the other is more fundamental: most people using these tools for slop only care about advertising money, they neither know nor care about how much of a difference artistic taste makes (hence also why there's so many YouTube shorts which are just a clip from a random TV show with inappropriate music added, sometimes as a mic-drop, sometimes just obscuring the dialogue, but how even are YT shorts monetised?) FWIW, there are models which are better than the majority of the YouTube slop I've seen*; but even then, when I show short clips from e.g. Google's "Flow" generator to others, those others find the results unimpressive. Also, the voice range is weird, my experience playing with them it had a much harder time of creating a decent working-class British accent than Dick Van Dyke. However, amongst the various issues Flow has, the most obvious is length, because even though Flow lets me "extend" a clip, it's not got enough coherency between each (8 second?) segment when I do. The second biggest issue is that, as with Stable Diffusion before it, it quite often produces stuff you just don't want. Junk dealers don't care about that and use the first result, an artist can afford to do 10 attempts and pick the right one, and will care to, too. That's something I can at least expect will be solved, or has been on non-free tools (control nets etc.) even though I don't expect slop dealers to ever bother with them. If each 8 seconds of video was actually Hollywood quality (they're not, in any sense including resolution) and they were coherent when extended (likewise, they're not), as there's 750 times 8 seconds in a 100 minute film, it would still be an existential threat to Hollywood if those segments cost $1333, because that's a million dollars and in the cost range of "micro-budget indie", rather than the tens of millions that even normal-indie films cost (before marketing etc.) or hundreds of millions that blockbusters cost. I have no idea how video generators scale up in cost with time or resolution, let alone quality, so I don't want to make guesses on how much extra compute (or indeed training data) would be needed for a coherent 2 minute cut-equivalent at 4k. As I don't like the epistemic collapse we're already witnessing with clips from video games being passed off as coverage of warfare, I very much hope this is one of those things which fails to scale up. * although even with early models, if you wield them as an artist rather than as a slop merchant you get interesting output; you may remember the Harry Potter Balenciaga videos from a few years back, same channel and you can see how things have changed: https://www.youtube.com/@demonflyingfox/videos https://www.youtube.com/@demonflyingfox/videos