James Cameron, the Oscar-winning director of films like Avatar, Terminator and Titanic, appears cautiously optimistic about the role generative AI can play in filmmaking, though even if he is wary of the “in the style of” prompts that have proliferated after images in the style of Studio Ghibli flooded the internet over the past few weeks.
“I think we should discourage the text prompt that says, ‘in the style of James Cameron,’ or ‘in the style of Zack Snyder,'” Cameron said on a podcast Wednesday, adding that “makes me a little bit queasy.”
And yet, Cameron acknowledges that the ability to create content that mimics great talents is undeniably interesting, and mirrors what he himself does in his own head.
“I aspire to be in the style of Ridley Scott, in the style of Stanley Kubrick. That’s my text prompt that runs in my head as a filmmaker,” Cameron said. “In the style of George Miller: Wide Lens, low, hauling ass, coming up into a tight close up. Yeah, I want to do that. I know my influences. Everybody knows their influences.”
Cameron was a guest on Boz to the Future, the podcast hosted by Andrew Bosworth, the CTO of the the giant Meta. The latest episode, which dropped Wednesday, featured an extensive conversation about generative AI, with Cameron sounding optimistic about its use in special effects, and uncertain about whether studios, tech giants and legislators should focus on regulating the inputs to the AI models, or the outputs.
Cameron, of course, joined the board of AI firm Stability AI last year. Stability is the company behind the Stable Diffusion image model.
“In the old days, I would have founded a company to figure it out. I’ve learned maybe that’s not the best way to do it. So I thought, all right, I’ll join the board of a good, competitive company that’s got a good track record,” Cameron said of the decision. “My goal was not necessarily make a shit pile of money. The goal was to understand the space, to understand what’s on the minds of the developers. What are they targeting? What’s their development cycle? How much resources you have to throw at it to create a new model that does a purpose built thing, and my goal was to try to integrate it into a VFX workflow.
“And it’s not just hypothetical, if we want to continue to see the kinds of movies that I’ve always loved and that I like to make and that I will go to see — Call it Dune, Dune Two something like that, or one of my films, or big effects-heavy, CG-heavy films — we’ve got to figure out how to cut the cost of that in half,” he continued. “Now that’s not about laying off half the staff and at the effects company. That’s about doubling their speed to completion on a given shot, so your cadence is faster and your throughput cycle is faster, and artists get to move on and do other cool things and then other cool things, right? That’s my sort of vision for that.”
When it comes to the controversial question of “training” AI models, Cameron seemed to suggest that regulators and lawyers should be more focused on the output of AI programns and tech, rather than the inputs and training data.
“A lot of the a lot of the hesitation in Hollywood and entertainment in general, are issues of the source material for the training data, and who deserves what, and copyright protection and all that sort of thing. I think people are looking at it all wrong,” Cameron told Bosworth. “I’m an artist. Anybody that’s an artist, anybody that’s a human being, is a model. You’re a model already, you’ve got a three and a half pound meat computer.
“We’re models moving through space and time and reacting based on on the our training data,” he continued. “So my point is, as a screenwriter, as a filmmaker, if I exactly copy Star Wars, I’ll get sued. Actually, I won’t even get that far. Everybody’ll say, ‘hey, it’s too much like Star Wars, we’re going to get sued now.’ I won’t even get the money. And as a screenwriter, you have a kind of built in ethical filter that says, ‘I know my sources, I know what I liked, I know what I’m emulating.’ I also know that I have to move it far enough away that it’s my own independent creation. So I think the whole thing needs to be managed from a legal perspective, as to what’s the output, not what’s the input. You can’t control my input, you can’t tell me what to view and what to see and where to go. My input is whatever I choose it to be, and whatever has accumulated throughout my life. My output, every script I write, should be judged on whether it’s too close, too plagiaristic, whatever.”
Instead, Cameron outlines a vision where more focused AI products help filmmakers create their visions more fully, arguing that AI giants like OpenAI and, yes, even Meta, are not really competing for Hollywood’s business.
“You look at OpenAI, their goal is not to make gen AI movies. I mean, we’re a little wart on their butt, I mean in terms of the scale you’re talking about, they want to make consumer products for 8 billion people,” Cameron said. “And I’m sure Meta is very much the same … movies are just a little tiny application, a tiny use case. That’s the problem. So it’s going to be smaller, sort of boutique-type Gen AI developer groups that I can get the attention of and say, ‘hey, I’ve got a problem here. It’s called rotoscope.'”
Cameron is currently working on the next Avatar film, Avatar: Fire and Ash, which will be released in December, and reportedly will include a title card noting that no gen AI was used in creating the film.
Ok, but the No Fakes Act seeks to advance creators’ rights by protecting their voices and likenesses from the unauthorized creation and use of digital replicas.
Cameron is talking about integrating the use of AI into the VFX workflow.
Ok, but the No Fakes Act seeks to advance creators’ rights by protecting their voices and likenesses from the unauthorized creation and use of digital replicas.
Cameron is talking about integrating the use of AI into the VFX workflow.
That would still need to mine. Just depends on if they add safe guards or not. I would think the actor would need to approve it. Because they are never going to get approval from the original artist.
That would still need to mine. Just depends on if they add safe guards or not.
He touched on that:
Quote:
“A lot of the a lot of the hesitation in Hollywood and entertainment in general, are issues of the source material for the training data, and who deserves what, and copyright protection and all that sort of thing. I think people are looking at it all wrong,” Cameron told Bosworth. “I’m an artist. Anybody that’s an artist, anybody that’s a human being, is a model. You’re a model already, you’ve got a three and a half pound meat computer.
“We’re models moving through space and time and reacting based on on the our training data,” he continued. “So my point is, as a screenwriter, as a filmmaker, if I exactly copy Star Wars, I’ll get sued. Actually, I won’t even get that far. Everybody’ll say, ‘hey, it’s too much like Star Wars, we’re going to get sued now.’ I won’t even get the money. And as a screenwriter, you have a kind of built in ethical filter that says, ‘I know my sources, I know what I liked, I know what I’m emulating.’ I also know that I have to move it far enough away that it’s my own independent creation. So I think the whole thing needs to be managed from a legal perspective, as to what’s the output, not what’s the input. You can’t control my input, you can’t tell me what to view and what to see and where to go. My input is whatever I choose it to be, and whatever has accumulated throughout my life. My output, every script I write, should be judged on whether it’s too close, too plagiaristic, whatever.”
Quote:
Originally Posted by mwynn
I would think the actor would need to approve it. Because they are never going to get approval from the original artist.
They already have protocols for when they use digital doubles, even without AI being a factor.
But mostly I think he's talking about using AI to help out with VFX work that's not necessarily human likenesses. Things like spaceships, or dragons, or dinosaurs.
But mostly I think he's talking about using AI to help out with VFX work that's not necessarily human likenesses. Things like spaceships, or dragons, or dinosaurs.
Yeah he is dancing on a minefield with that statement. We are constructs. We are not perfect clones.
Let see what happens when an AI tries to create something living. I have a feeling it will be like The Fly.
The specific example of Dune 2 he gives is interesting, because the VFX artists did use machine learning to help automate the Fremen eye-glow effects. That's a direct case where AI helped save time and money so artists could focus on more interesting tasks. I also have a theory that part of the reason Cameron's "4K" "remasters" used 2K sources and (in the case of Titanic and Avatar 1) added interpolated HFR is that he was using them as a test bed for ways to save time on Avatar sequels. I can imagine him getting frustrated with the longer render times needed for a 4K, 48fps movie and thinking it would be faster to do a 2K, 24fps render and upscale it instead. There are certainly applications for machine learning in VFX, but it's a real slippery slope between "automated rotoscoping and improved upscaling" and "replacing actual artists with AI slop."
There are certainly applications for machine learning in VFX, but it's a real slippery slope between "automated rotoscoping and improved upscaling" and "replacing actual artists with AI slop."
The specific example of Dune 2 he gives is interesting, because the VFX artists did use machine learning to help automate the Fremen eye-glow effects. That's a direct case where AI helped save time and money so artists could focus on more interesting tasks. I also have a theory that part of the reason Cameron's "4K" "remasters" used 2K sources and (in the case of Titanic and Avatar 1) added interpolated HFR is that he was using them as a test bed for ways to save time on Avatar sequels. I can imagine him getting frustrated with the longer render times needed for a 4K, 48fps movie and thinking it would be faster to do a 2K, 24fps render and upscale it instead. There are certainly applications for machine learning in VFX, but it's a real slippery slope between "automated rotoscoping and improved upscaling" and "replacing actual artists with AI slop."
While I think this could fit well in the AI Hollywood threat thread, Cameron is a major director who influences the industry, so I can understand its own thread.
I'll listen to the podcast, but I'm guessing as long as he creates all of the initial assets, then programs the AI to do the in between work, that would save a lot of money and time. Like when you create a painting, and the AI animates it with a panning camera. That's amazing technology there.
Or how you photograph 4 angles of a car, and now you can use it as an animation model with AI in gaming or movies. Computer animation is a primitive version of AI, in that the computer does all of the in-betweening animation for the artist who mostly just sets the key frames and curves of animation, along with creating the model character rig.
On the other hand, AI that steals from artists randomly online is not good IMO unless it gives those artists credit, or doesn't make a dime from it.
By democratizing filmmaking and allowing people to create movies with no actors or even sets, just a script, AI is going to do to film what protools and the internet did to music when they removed the high cost barriers to recording and distribution in the 2000s. By significantly dropping the financial barriers, the market is going to be flooded with content, some great, some mediocre, but most awful. On top of this, films will become hyper-microtargeted to specific subcultures and everything will fracture into niches, with indie films that break into the mainstream being rare. There will still be great content out there, but it's going to be way more time-consuming to find it.
That's how I saw the 2010s indie music scene. I see it as maybe the best decade for music since the 1970s as online distribution allowed an explosion of creativity with no barriers. The only problems were (1) it was really hard to find the good stuff in the sea of mediocrity and most people just don't have the time to search it out; and (2) while a tiny few bands were able to make a semi-OK living, it's really, really hard to make a living in the indie music scene if you aren't already wealthy.
By democratizing filmmaking and allowing people to create movies with no actors or even sets, just a script, AI is going to do to film what protools and the internet did to music when they removed the high cost barriers to recording and distribution in the 2000s. By significantly dropping the financial barriers, the market is going to be flooded with content, some great, some mediocre, but most awful. On top of this, films will become hyper-microtargeted to specific subcultures and everything will fracture into niches, with indie films that break into the mainstream being rare. There will still be great content out there, but it's going to be way more time-consuming to find it.
That's how I saw the 2010s indie music scene. I see it as maybe the best decade for music since the 1970s as online distribution allowed an explosion of creativity with no barriers. The only problems were (1) it was really hard to find the good stuff in the sea of mediocrity and most people just don't have the time to search it out; and (2) while a tiny few bands were able to make a semi-OK living, it's really, really hard to make a living in the indie music scene if you aren't already wealthy.
Good points about how flooding of new content (new movies or music or videos) means a lot will get lost in the waters passing by.
The music industry has pretty much dissolved for smaller bands. I've seen once famous bands releasing their own albums instead of through major labels (many of which probably shut down) and their album goes nowhere surprisingly other than with their diehard fans. I hear touring is the better way to earn money as a musician. Or to be part of a compilation or other form of entertainment/soundtrack, etc. so it can be discovered as part of a larger block in the flood waters.
Good points about how flooding of new content (new movies or music or videos) means a lot will get lost in the waters passing by.
The music industry has pretty much dissolved for smaller bands. I've seen once famous bands releasing their own albums instead of through major labels (many of which probably shut down) and their album goes nowhere surprisingly other than with their diehard fans. I hear touring is the better way to earn money as a musician. Or to be part of a compilation or other form of entertainment/soundtrack, etc. so it can be discovered as part of a larger block in the flood waters.
The major labels are all still there. An indie band can't just call up Clear Channel or Audacy and ask them to play their song on Pop Top 40 radio.
The major labels are very good at signing up any artist/band who gets white hot traction from a single, then distributing it to Spotify or Apple Music where the apps payout $16 to the artist for a billion streams after the label takes their agreed-to cut.
Then the major label will let the artist go on tour to make a living, only the label is leasing the artist's songs back to the artist when they play them live.
All countries
Argentina
Austria
Belgium
Brazil
Bulgaria
Chile
China
Colombia
Cyprus
Czech Republic
Estonia
Denmark
Finland
Greece
Hong Kong
Hungary
Iceland
Indonesia
Ireland
Israel
India
Latvia
Lithuania
Malaysia
New Zealand
Norway
Philippines
Poland
Portugal
Romania
Russia
Singapore
South Africa
South Korea
Sweden
Switzerland
Taiwan
Thailand
Turkey
Ukraine
United Arab Emirates
James Cameron says AI could cut cost of blockbusters in half
Linear Mode
