Why Hollywood and gaming struggle with AI

Innovator's dilemma, legal pushback, creatives with pitchforks, and more

You'd think that AI would drive a boom for the entertainment and video game industry. Instead, incumbents from entertainment are facing incredible friction as they attempt to experiment with generative AI. Of course there is a positive note — this opens up a huge window for startups to reinvent entertainment and gaming as a result!

Why are big cos in entertainment + gaming not aggressively embracing AI? And why does that turn into a startup opportunity? There’s a few key reasons:

• If it works, don't fix it

• Desire to "bolt on" AI as a feature rather than AI-first

• Legal pushback from questions on IP ownership

• Pushback from creatives/designers

• It's hard to hire AI engineers

• The silver lining: The $1000 blockbuster film

Let's explore each of these.

If it ain’t broke…
Many of you know that I moved to LA in recent years, and I often speak to prominent folks in the entertainment + video game industries. The good news is that these executives are very sharp, and are following all the technology evolutions. They are both hugely optimistic and they acknowledge that AI will reinvent their industries. But there is no will to proactively embrace the technology. After all, when you have a series of franchises and delivery channels that work, why take the risk with new tech that might end up require a lot of rethinking of business models / workflows, might cause rebellion and dissension among customers/employees, and maybe not even work?

Within gaming, image that it’s taken over a decade for bigco developers to embrace mobile and free-to-play. Call of Duty finally released their first mobile game in 2019, alongside Blizzard's Diablo Immortal mobile game in 2022, both more than a decades after mobile was released. AI-first gaming will be similar. Or that Hollywood continued on analog film reels for decades after it was possible to film everything digitally, and only switched during COVID.

Bolting on
When new tech is embraced by incumbents, it's often "bolted on" rather than reinvented from scratch. That is, it's much easier to imagine what it's like to use AI in the backend production process to build games and films we already know and love. Both films and video games can cost $100s of millions to develop. There's interest for cost savings from the business side there, which is why there has been a pattern of outsourcing and offshoring. (We'll discuss later why the creative arms of these companies are up in arms).

However, AI will deliver more than cost savings. The biggest opportunities in the last revolution in content creation — user-generated content and Web 2.0, which gave us the world’s most popular mobile apps — grew by creating new genres. There has been a mountain of vlogging, unboxing videos, dance videos, product reviews, all done with a low-production selfie sticks, and that is what’s causing TV/film to slow in consumer engagement. These are brand new genres of content that haven't existed in the past. And the next jump in content creation — enabled by AI — will also create new genres. This will require experimentation and radical innovation, most likely coming from startups, not incumbents.

Timing plays a major role here. The gaming industry is in transformation, with a steady drumbeat of M&A changing the landscape, and a lot of RIFs as companies restructure. Hollywood has much the same happening, with a flat/shrinking film market transitioning to streaming. Who wants to take risk right now, and invest in breakthrough new genres?

Startups will embrace this opportunity. Big technology dislocations like the type we're undergoing right now typically create huge opportunities to build new, strong form, "native" products. For mobile this meant startup products like Uber and Instagram, which emerged several years after app store, versus weak form apps that happened right away, where people ported well understood apps like email or fart apps or flashlights. AI-native gaming and entertainment experiences are likely to deviate significantly from our current understanding of the medium.

They might be more meme-like and ephemeral, meant to be created in a few days and played for a week or two, alongside a major world event like the Super Bowl or a Presidential Election -- because there was a funny moment that became sort of a meme game. Or perhaps the game experiences themselves will target much smaller niche audiences, the same way that people often build websites for small groups. Perhaps there will be new genres, just as "6 second dance video" has become a new category. We'll likely see low-hanging fruit like AI sim games and narrative gaming -- anything with a lot of content, many characters and NPCs, end up being reinvented first.

Legal hurdles via training data
The other big obstacle for incumbent game+entertainment companies comes from the way AI works -- in particular the fact that the models are trained on huge amounts of text/image/video/3D assets that are presumably "public domain" but what if they're not? The legal departments within many of these companies have held their product teams back because sometimes the models seem to spit out copyrighted content when prompted with the right set of prompt text. On the other hand, startups tend to run towards this type of risk and validate product/market fit first, as YouTube did early on with all their pirated music video and TV content. The assumption from many startups is to try and prove that consumers love this kind of content first, and then to put themselves in a better legal position as they grow. This may or may not be a good idea, but it's another thing that makes AI-first startups must more productive.

And of course, all of this makes the companies also reluctant to share their data with AI companies because they might suspect their IP will eventually get incorporated within the model. We dealt with some variations of these trust issues with cloud computing, and in the end a lot of companies put their email+data in the cloud and everything's fine. Or maybe we'll have lots of "on prem" AI models trained on small bits of internal data + big external public stuff, and companies will try to keep things within the firewall.

The creatives push back
All of the IP in these gaming+entertainment companies are of course created by the vast numbers of creative workers within these companies -- the artists, writers, designers, and so forth. Maybe of these folks have a tough time grappling with the idea that these AI models are built from other peoples' work. Unlikely pure software companies, you might have a ratio of 3:1 artists:engineers within these organizations -- and thus the creative employee base is a very powerful one. There's already been examples, like the SAG-AFTRA strike, trying to head off AI as a disruptor in the industry. I'm sure we'll see more.

The most positive thing I can argue with the creative work force within these companies is to point them to the trajectory of the software engineer. Everything that's happened in the past decades -- open source, cloud computing, better languages, better IDEs, etc -- have made software engineers even more general and applicable to any problem. It's grown the demand for their skills, not reduced it. The world would not be better with 100,000 FORTRAN engineers -- it's better we have 10s of millions of devs that can tackle any field. I think that as creative workers are more empowered with better tools, we might see them able to develop entire prototypes or start companies themselves (no code!) and be otherwise even more valuable. And that might increase the demand for their skills. Similarly, you might just see them make more content, or try more things. All of these can be good outcomes -- not to say that there won't be winners or losers, but that some forms of creativity are likely to thrive.

The difficulty of hiring engineers
All of this is further compounded by the simple fact that it's hard to recruit great AI engineers. It's difficult for everyone, even startups and tech companies. Many of you have been following the fact that a top-flight AI researcher often makes millions of dollars a year. We’re not talking about a leader, or executive, but rather a rank-and-file AI PhD working on a commercial project. The tech industry has been seeing it for years, fighting over this type of profile for car autonomy projects, or for customized feed algos. But now it’s widespread.

It's hard enough for top-end tech companies that pay very well, or where there's significant equity upside. But it's even harder for older and more established companies that can't stretch using high-upside equity. If AI becomes as impactful as we all think it'll be, there's a major issue in simply getting good talent into incumbent companies. This sort of talent adverse selection will make it hard for more mature industries to compete. They’ll have to resort to vendors, with all the associates headaches of that.

Enabling the $1000 blockbuster movie
I want to end this discussion on a positive note, for what AI can do for the creative indistry. The positive way to frame all of this is the concept of the "$1000 blockbuster movie." Today millions of dollars (and hundreds of people) are needed to make a full length movie, a multiplayer game, or a multi-season TV series. But compare that to a humble book...

Writing a book just needs a laptop, coffee and lots of time. Not much money. Add up the laptop and all the coffees, and you maaaaybe spend $1000. What if an individual could make a full-length summer blockbuster movie by themselves with just a laptop and some coffee too? It would be 100,000x leverage to be able to author a movie (or game, or TV series) - the same way that someone can write a book It would require lots of effort and grit. It's not easy to write a good book - no diff for a great movie. But it would be possible, when it's not today.

What an explosion of content this would create! In every form. I think of it as the next phase of lowering the cost of producing and publishing content online, the way that YouTube, Instagram, and other user-generated content platforms have democratized distribution. Many of us who are following the lightning fast developments in generative AI intuitively know that the end outcome of these tools will change the process of creative work forever. This will increase by 10000000%s who can create movies, or books, or otherwise. And make the content better as geniuses can increase their output. Rather than taking jobs, this will create more opportunities for creatives to color outside the lines of their own medium, and work in adjacent forms of media. An author will build a videogame around their work. A game designer will create an award-winning TV series in their game’s world. It will cause an explosion of entertainment and content and art — something that’s good for the world.

But we have a long way to go. Not every incumbent entertainment or video game company will have the ability to succeed. After all, the long list of problems listed in this essay are all interlocking issues. This is why it's super hard for incumbents in entertainment and gaming to embrace AI. Although it's the obvious New New Thing, many organizations are fundamentally limited by their internal structure, legal liability, their desire to incrementally build their businesses, etc.

If it’s hard for the incumbents, then let’s the startups thrive. New companies will revolutionize entertainment, and consumers will benefit. Much more to come.