Protest, cooperate or embrace - AI vs. Hollywood

If you spend over 30 seconds on your favourite news or social media sites, you will see a negative article or post about artificial intelligence. Following the announcement of the Screen Actors Guild (SAG) strikes, which are partially blamed on a broad-based distrust and fear of AI being used in the film production process, you wouldn't be wrong if the image starts to feel familiar. 

The fear of A-list actors at the front of the picket lines is understandable. Is it easier to create Non-Actor-Characters in scenes rendered mainly in Unreal 5? Yes. Was machine learning used in an extensive scene in the latest Indiana Jones movie to de-age Harrison Ford by forty years? Yes, and he is still on strike. 

Are generative AI tools used for concept art and to create random names for alien planets? Yes. 

What seems to be concerning SAG is Hollywood’s ability to control the usage of AI in their industry. The fear generated by things moving so fast that they seem out of control is a potent emotion. But herein lies the issue that has been revealed countless times when new technologies have been introduced. You will struggle to control it. 

As President Truman said on the 3rd of October in an address to Congress on atomic energy, “constitutes a new force too revolutionary to consider in the framework of old ideas”.

He suggested fostering “conditions under which cooperation might replace rivalry in the field of atomic power”. So, let’s consider this wave of artificial intelligence development and how it might impact our entertainment industry. Will striking and halting the creation of unionised actor/writer-driven content build audience support or distaste? How could the entertainment industry “cooperate” instead of opposing this technological advance?

We respect creative’s rights to preserve their likeness and work, but here is why seeking a blanket ban on AI is unlikely to be a fruitful exercise:

Suing the leading companies in Generative AI will be highly complicated.

Generative AI platforms are trained using broadly publicly available data in the public domain or within mass public digital communities. For artists to stop Gen AI products from scouring GitHub, Wikipedia or arXiv for videos and images showing their face or work, it will require a lot of investment up and down the value chain. 

Take the popular AI tool, Stable Diffusion, as an example. 

Stable Diffusion is a free trial text-to-image AI product that uses a large language model (LLM) to generate images from natural language descriptions. Stable Diffusion grants all images made as part of the free trial the Creative Commons licence, so no one technically owns them, and anyone can use them for just about anything.

 It is trained using data primarily sourced from massive datasets from the non-profit LAION. LAION’s datasets are compiled by another non-profit called Common Crawl, which scrapes billions of web pages monthly to aid academic web research. 

As the Atlantic more eloquently put it, “Stable Diffusion helped fund a nonprofit and gained access to a dataset that is largely compiled using a different academic nonprofit's dataset, to build out a commercial tech product that takes billions of randomly gathered images (many of them the creative work of artists) and turns them into art that may be used to replace those artists traditionally commissioned artwork.” So, if you are a creative, you can prevent Common Crawl from scraping your site.

But many of those images originate from sites like Pinterest, where others upload and share the content. It's unclear how an artist could stop Common Crawl from scraping Pinterest.

As I am trying to articulate, this isn’t so much of a rabbit hole; it is a bloody warren!

The complexity, cost and scale of the litigation and regime change to enact equitable outcomes for artists are likely insurmountable. Not even seven-figure donations to SAG AFTRA from The Rock could tackle this problem. 

The can of worms surrounding the attribution of value from IP could have immense implications, and it may not be the can that SAG AFTRA wants to crack open. 

This year's most prominent box office performers have been based on character IP.

Do people pay over $109 for a family to watch Super Mario because Chris Pratt and Seth Rogen provide the voices for Mario and Donkey Kong? I would argue not (there was much controversy about Pratt’s appointment). The days of the Michael Ovitz-driven star-packaged flick that led with its lead actor over perhaps the IP is probably long over. With this in mind, what percentage of the streaming revenue that HBO generated from The Last of Us should be attributable to Pedro Pascal? Or should it instead be more attributable to the dozens of artists who created the initial model and rig for Joel and to Troy Baker for the original voice acting? 

The can of worms surrounding the attribution of value from IP could have immense implications, and it may not be the can that SAG AFTRA wants to crack open. 

SAG AFTRA wants a more significant share of the streaming platform’s revenue, but the kitty hasn’t much left. 

Perhaps, we need to look at the state of the companies that pay actors and writers salaries and honour their residual rights. As the summary from the Hollywood Reporter shows, it isn’t rosy in the financials for most of the largest studios in the world:

The very companies they are striking against are the ones with the most margin (and it isn’t much!), growing subscriber base and an extensive backlog of content to keep their customers content. The last strike created many reality TV shows, which did not require the services of unionised talent, and helped create the Celebrity Apprentice (perhaps even why Donald Trump gained the audience to become President!). Without wanting to be political, the ramifications and ongoing reverberations generated by mass industry strikes can be alarming and impactful in ways we never imagined. 

We shouldn't forget the fallout of these strikes on the industry and the livelihoods it supports. 

Looking locally at New Zealand, several major productions being shot in our country have been cancelled. The movie adaption of Minecraft, starring Jason Momoa, has been put on hold, laying off over 700 people from the set. The impact of this occurring during a recession is significant and should not be underestimated. 

To put these strikes into context, there has been entertainment industry strikes approximately every decade for almost a century as new mediums succeed each other, from television to cable to VHS to mobile to, now, streaming. SAG performers last went on strike in 1980, while screenwriters did most recently from late 2007 to early 2008. 

Every new kind of technology creates not only different arrangements of labour and management but new forms of content that become popular and earn revenue.

The fact that the unions are talking about AI potentially indicates that the unions are not just concerned about the present moment but also trying to establish assurances about the near future.

I believe that contracts in the future — or future strikes in protest of them — will continue to reflect technological change.

I think that actors may walk away from these strikes with a better deal, whether a great deal will keep them satisfied for years to come; I still wonder. When it comes to tackling the issue of AI then, I believe these current strikes will not solve this. I think it is part of a more comprehensive societal reckoning that humans need to have with a technology that is too significant to be encroached upon with strikes and militancy. 

If a part of President Truman’s legacy is that cooperation replaces, rivalry could be wishful thinking. 

Who might emerge stronger from this? 

If the strikes continue into the end of Q4 2023, we expect more significant interest in international content not affected by the strikes, perhaps from the powerhouses of Nigeria and India. Equally, with many VFX studios affected by the halt of significant productions, they may redirect their teams into creating their IP and the next Illumination Studios or Pixar. The growth of content that does not require voice actors may rise and become a new genre of silent movies!

You might pull out from this piece that I am taking a somewhat antagonistic stance towards the rights of actors, but this is not my intention. As Christopher Nolan recently commented in Variety, “One of the best casting choices in film history, in my opinion, was when [Jon] Favreau had the vision to cast him as Iron Man. And you examine what happened and the direction that everything took as a result. And I believe Jon knew exactly how amazing of an actor and how amazing of a promise Downey had. Then the allure of a movie star, that amazing magnetism, enters the picture.”

An algorithm cannot predict or replicate the chemistry between iconic actors, directors and the IP they portray. The serendipity of an emerging director meeting with a down-on-their-luck actor and casting them as an obscure comic book character that goes on to spawn a franchise worth billions of dollars is pure magic. 

I would like to give you a portrayal of artificial intelligence in the entertainment industry where it isn’t a device that steals someone’s identity or displaces the entire human cast. How about looking at AI as a platform to augment human output instead? A writer could work on significantly more projects than before using machine learning and earn more money. Maybe it reduces the time spent coming up with names for obscure planets or aliens in the Star Wars universe or automatically identifying plotholes that the human eye might have missed. 

Here is another bit of futurology for you - imagine a time in the near future when you can pause your favourite scene in a show, zoom in on a character actor or extra in the background (digital or human) and then ask for their back story. A generative AI platform using predetermined prompts from the writer creates a captivating story that expands the universe and entertains the viewer even more. 

Suppose actors continue to be disillusioned by machine learning-driven makeup or de-ageing. In that case, the actors may go back to kickstart a stint on Broadway or the West End of London and create IP from in-person, live performances. The exceptional quality of the Royal Shakespeare Society actors has benefited from years of treading the boards in front of live audiences in magnificent theatres. I would pay a high price to see Jeremy Strong, Timothee Chalamet and Cate Blanchett perfecting their art live.

Conclusion

Fear breeds confusion and the acceptance of mistruths. In the case of AI and the entertainment industry, this technology has the potential to delight audiences and fill their thirst for ever-increasingly rich worlds of content and characters. 

Yes, I can foresee a sector that employs hundreds of human extras to bring scenes to life when they could be rendered and directed automatically through a combination of AI and existing rendering engines. Directors have been doing this for decades using CGI, and I don’t see any reason to stop this. It is also a more sustainable way to create sets and potentially hit a net carbon movie shoot. 

AI will never replace art. It may augment it but never replace it. Cooperation would be a far more efficient solution than simply denying its existence. 

Here’s to finding a middle ground on revenue share from the streamers and studios, getting everyone back into employment, and delighting audiences again.