Where will the AI Horde strike next? AI video, social media, and Hollywood

1000s of AI startups ride into entertainment, among dozens of industries

Look at this graph, illustrating the rise and fall of digital cameras, and their collapse triggered by the ubiquity of smartphones. This graph is incredible of course, and it captures the simple story of technological disruption in one visual. It’s so easy! Just don’t get disrupted.

Of course it’s apt to start an essay about the coming AI disruption with a chart like this, but honestly it’s such a cliche. Duh. Boring. We all got it. After all, everyone has read the same business books, listens to the same podcasts, and knows all the same jargon and the same predictions. In fact, all the these legacy incumbent industries these days they are staffed with a slew of tech-forward millennials and MBAs who have all read the relevant Harvard Business case studies.

We know AI disruption is coming for many industries. The AI Horde, consisting of thousands of VC-backed AI startups, are starting to go after every industry as a general computing technology that can replace and augment labor. So what happens next, as they compete with incumbents?

“AI will disrupt Hollywood!!!” — really? will it?
To take an example of a single industry, scroll enough on social media and you’ll see a prediction like this: “AI will disrupt Hollywood!!!” This kind of assertion is usually accompanied with an amazing 30 second video that gives a taste of the latest in AI — the output of a bunch of video generation models, speech synthesis, and AI music, all cut in the way that is highly entertaining for 30 seconds. And the prediction mostly kinda feels right. And it is genuinely amazing how fast gen AI video models are progressing — today the output is still a little janky, but the rate of improvement is incredible.

We’re going to dig into this specific example of AI video and Hollywood/YouTube/etc, but this is being said for many industries that are thinking about AI. And I want to give the AI startup’s view, not the view of the incumbent. This is a different approach because the disruption theory, and the viewpoint presented in the Innovator’s Dilemma is focused on how an incumbent should best react and protect a well-defined product category. The POV makes sense because most management theory exists to perpetuate the incumbents, and the MBAs are hired to middle manage big, enduring conglomerates.

Contrast that to brand new startup, which has a radically different POV. The startup is just trying to survive. They are trying to find/scale product market fit, and navigate a complex system of niches and markets trying trying to find a foothold in at least one, so that they can continue to get more funding and continue their journey.

The incumbent thinks defense, and is trying to defend their particular castle. They ask, how do I protect my specific position in my industry while technologically disruptive AI startups enter it?

The AI startups think offense and acts like a barbarian AI Horde, exploring and probing and retreating, trying to figure out what will work among many possible targets. They’ll attack any castle that feels weak, and switch to another if it doesn’t look easy. They ask, if I can build products with this new tech, which market should I go after where competitors are weak? Should I double down where I am, or pivot to attack another? Should I attack the incumbents directly, or try to sell tools to them? Do I go premium and white glove, or build something simpler and low-end?

This is why the AI versus Hollywood situation is so interesting.

The big choices for AI video
The AI Horde’s upcoming decisions are complex. Let’s assert that the AI video gen tech is just going to get there eventually, and we’ll soon be appreciating video in many formats. Ahead of this inevitable outcome, let’s say you are a skilled new startup team trying to best take advantage of this upcoming trend.

The multitude of decisions might start like this:

• Should we work within the existing ecosystem?

  • Sell tools to existing entertainment companies

  • Create film/TV and sell the content (“Pixar for AI”)

  • Buy an existing film/TV studio and vertically integrate AI

• Or, should we target the social media ecosystem?

  • Sell tools to creators

  • Become a new kind of creator that uses AI to create a lot of content

  • Play some other role in the existing ecosystem (ex: build AI video etc)

• Or, build something new that exists independently?

  • Create a new app that uses AI to disrupt an existing category (like AI ReelShort)

  • Innovate on a new AI-first format (interactive? companions? a new kind of game-like experience?)

  • … and much more

These are a few of the big choices, but really there are an infinite number of decisions and sub-decisions. We could add even more detail by sitting on beanbags and brainstorming more, but this is a good place to start.

Imagine yourself as a new AI startup trying to figure out where to attack. There’s castles defending every bulleted line on the list, which is on one hand very bad, since now everyone has read all the same books and folks are rapidly writing long memos about the threat of AI. But on the other hand, the entire advantage of being part of the AI Horde is that you’re small and fast and can attack wherever is weak. And barring that, many of these bullets are B2B and you can just arm the big guys with weapons to fend off the AI horde.

Work with the incumbents? Weak-form vs strong-form?
Depending on your interests and prior background, you might have an attraction or an aversion to interacting with players in the incumbent industry — for the sake of this example, we’re talking about Hollywood specifically. (I find myself mostly disliking startup ideas that depend on these incumbents to adopt technology, since they have a long history and culture of avoiding it whenever they can, which I’ve discussed here: Why Hollywood and gaming struggle with AI). But it does come with one major advantage, which is that they have money and motivation. They also have well-known and loved brands, characters, and franchises. Theoretically a new AI startup could cut some amazing deals and have a jumpstart on their journey.

So here are some ways to approach working with Hollywood:

• Pixar for AI aka a new AI Hollywood studio: If an AI startup can hybridize with a studio, then theoretically one could create the next major film idea, sell it to a major studio (Sony, Paramount, Warner Bros, and the like) and receive financing to make the film without raising additional venture capital. This would generate revenue, and the AI studio startup could continue with a series of these, and if one is a hit, it might become a very cost efficient way to get going. This is, plus or minus, the story of Pixar, Toy Story, and Disney of course. The biggest challenge here is that you’re required from a tech perspective to build something at the highest quality level for the silver screen right off the bat, and I don’t think Clayton Christensen would approve nor would it be the easiest thing to do. But this option is there. My colleague Jon Lai wrote an essay on this whole approach (and more) and it’s worth reading. One interesting side question here is whether you’d rather back the luminary Hollywood team who then goes and adopts tech, or would you rather fund a tech founder who then sells into the industry? Both are hard, but again, depends on your tastes. Many tech VCs would rather back the tech founder, but the dream of course is Ed Catmull with John Lasseter.

• AI-enabled/tech-enabled studio: There’s another interesting approach here which is to actually buy a production studio that already has a business, team, and process, and create an AI team working alongside them to aggressive build/buy/partner on all the tech needed. We’re seeing this attempted in entertainment already, but also fields like legal, customer support, accounting, and so on. This might actually be easier to do than to try to sell software.

• Sell tools: There are a ton of workflows within the film industry that could use AI tools today. Some are obvious, like dubbing, internationalization, etc. These are nice niches that might be good starting points. Or you could look at other issues, like digital asset management or video search. Or editing, including the ability to fill in gaps of video with AI-generated content, rather than reshooting. The failure case for all of this is that you’re selling tech into companies that have tight budgets and historically have not been great customers. And maybe your customers also hate AI. But on the other hand, if you can make people more productive and/or your new AI app works on an outcomes basis (like the total project fee to dub a film, rather than as a $/seat tool) then that might work too. But also might be hard.

All of this might work, but there’s a major problem: There are major constituents within Hollywood that absolutely hate AI. Given that there’s so much pushback from unions and creatives against AI, will this approach even work? Maybe major studios will object to all uses of AI tools placate their creatives, and that means they won’t use tooling, won’t engage with companies that use AI, and won’t buy content that was created with AI. Even if it’s good for them financially, one of the big differences between the entertainment industry and the software industry is the former has very strong union control, and strikes can cripple an entire year of films. If you’re a pessimist, this is your take, and that’s why Hollywood might take a long time (like way beyond what’s reasonable) to adopt AI, and instead all the AI storytelling will happen on social media and other freer platforms.

The optimist’s POV on this is to observe that actually there is a battle between the business people and creatives in the entertainment industry — the former actually know their whole business model is broken, and are excited about AI vastly increasing quantity, promoting more creativity, and reducing costs. So they will find a way. And also that Hollywood itself is very fragmented with a zillion smaller production studios and companies, many of which are excited about AI. (We’ve been pitched by many!) These techno optimist studios are bound to do amazing artistic work, maybe initially with a lot of human-in-the-loop, and as a result will open doors. Some of the major distributors like Netflix and Amazon might be less affected by union issues, so perhaps they could buy a new AI film and open the floodgates. Or perhaps smaller distributors will prove things out in other geographies, show success there, then bring it to the US as the Overton window shifts.

Building for the digital-natives
If you decide not to build for Hollywood, then you are building for the Internet, which might mean social media or also potentially just building your own app. This could manifest in a bunch of different ways:

• AI Cocomelon: One version is to create a next-gen media network like Cocomelon (if you haven’t heard, a very valuable kids network sold for $3B recently) or Mr Beast or something else similar, but with AI-generated content. These advantage here is you don’t need anyone to green-light the content the way Hollywood works — if people watch it, you will make money. However, the business model is a residual stream of advertising against the videos over a long period of time, rather than an upfront financing that helps defray the cost of building the content. Thus, initially content in this format will look cheap/low-quality, and they need to build a huge portfolio to generate the stream of capital needed to be successful. This is also where AI video gen startups creating anime or animation (kids/adult) will be great because this tech is working well enough right now — you just need the creative talent in there to create new stories.

• Tools for social media creators: We are already seeing plenty of these, and it’s sort of a “sell picks and shovels” — helping creators with video tools that allow them to use AI to create new content. This is potentially a very large market since billions of people now use social media and most of them create content of one form or another. However, obviously very competitive and many utilitarian products lack defensibility. (Btw, I personally love Captions (an a16z startup) which has really pioneered this space!) But perhaps these tools will undoubtably become popular and am sure there will be multiple billion dollar companies here

• New app that’s an AI ReelShort/TikTok/Twitch or other video type: You could build a new app, but have it centered around the output of various AI video models to target one of the many video formats that are already out there, for example, an AI TikTok app, where the model generates engaging 30 second videos, or AI ReelShort, which is a micro-transactions driven narrative content app. Or an AI Twitch, AI OnlyFans, or something else entirely. These are all new standalone apps that are fast-following existing video formats. Perhaps the user doesn’t even know they are AI generated.

• Real-time interactive content, gaming, and other new stuff: The creation cycle of Film/TV works the way it does because there’s an impossibly long loop between shooting the content, editing the content, versus watching the content. If you can generate video in real-time, then the loop works in a completely different way. Perhaps you generate the video all on the fly, and the story all real-time as well. This makes sense particularly in a world of gaming where people want to make choices and see the world (and characters) react around them. Perhaps there is something between a game and a film. Perhaps there are games that incorporate 10x the cinematic content, and is its own genre. It’s hard to say what might live in this space of next gen interactive content, but this is probably one of the most interesting opportunities since it’s where “AI native content” actually lives.

Of course there are more ideas beyond this, but I’m cheating here a bit and saying “new formats” are a placeholder for all the cool new things we’re bound to see. The digital-native approaches are highly appealing for Hollywood outsiders, who don’t have the relationships within entertainment. It’s also appealing because you can create low-fidelity content — perhaps meme content, or shorts — and gain traction. It doesn’t require you to jump directly to what can be shown on the silver screen right away, which is a huge win. But on the other hand, all the major problems that tech startups face exist here — scarce financing, expensive marketing channels, difficult distribution for new apps, etc.

Whew, so we’ve talked about a lot of permutations now. A new AI startup might look at the maze ahead, with all the potential options and winding paths, and inevitably be faced with confusion. Ultimately that’s OK — building a startup is about thinking probabilistically, launching and iterating and learning and launching again. You might start by wanting to sell a new animated series to Hollywood, but then realize you’d rather try to sell the tool. But perhaps the customer base is slow to buy, so you release a self-serve version that’s then embraced by social media creators. You can quickly pivot from one part of the maze to the other.

Here’s why disruption theory is so limiting when you look at the view of the world from the incumbent, agonizing if the AI Horde will go after your particular market. The reality is that the Horde is moving as a wave, exploring every niche, and every new technology. They are hard to defend because it’s hard to even tell which one is actually going after you — is it the AI short film app that sucks away all your consumer time, thus tanking your streaming subscription revenue? Or is it the AI tools company that enables all your competitors to create amazing 3D animated content as you’ve been doing for decades, eroding your defensibility? The AI Horde does it all.

How this applies to other categories and AI startups
To wrap up here, I’ve spent a bunch of time talking about Hollywood and the upcoming AI video gen models but this discussion could be applied to many other markets. You often face some of the same choices:

• Do you work within the existing market, or try to create a new one?

• Is it better to sell tools, or to go directly to consumers?

• Can you accelerate by buying an existing player, and adding AI? Or do you start a new firm? Do you start with technologists, and add domain experts, or the other way around?

You could be talking about many industries here — marketing/PR, accounting, consulting, legal, customer service. And my guess is, multiple approaches might end up working.

So I want to go back to the original statement from the beginning of the essay: “AI will disrupt Hollywood!!!” — will it? I think AI will disrupt, but also be adopted and assimilated. There will be new consumption experiences that compete directly with film/TV, but also ones that indirectly compete. It’ll all happen.

After all, if you went back in time 20 years and realized that online content would have been big, you might have thought: Let’s go strong form. People will just watch internet content, don’t worry about where the incumbent media formats will go. Yet both Netflix and YouTube worked. You’d ideally have bet on both. I think the same thing will work. The AI Horde is so powerful that we’ll see it reinvent Hollywood, gaming, YouTube, and more. It’ll affect the content, yes, but also AI tools, and AI-enabled companies, and much more.