AI Search
The Sora model text-to-video is being displayed on a smartphone with the OpenAI logo visible in the background in this photo illustration in Brussels, Belgium, on February 16, 2024. (Photo Illustration byJonathan Raa/NurPhoto via Getty Images)
OpenAI’s retreat from AI video is a reminder that media success still depends on the things technology cannot easily manufacture: distribution, storytelling talent, shared cultural relevance and a business model that actually works.
Making content and building a media business are about as similar as owning a Peloton and qualifying for the Tour de France.
We know the story: New technology arrives; social media fills up with awe; executives start fantasizing about transformation; and a lot of people quietly assume that because the tools are getting radically better, the business model must be getting stronger too.
OpenAI’s retreat around Sora is very instructive, but it’s not necessarily a surprise – especially the part of its deal with Disney. It is a story about a broader misunderstanding of what actually makes entertainment work, because it requires a lot more than stunning technology.
There are four things people forget when it comes to content creation:
Distribution matters a ton and there will always be an advantage to those who have that pipeline. Being a content creator is hard precisely because few have that access.
This is where a lot of AI optimism runs into the wall of media economics. The internet dramatically lowered the cost of producing content, but it did not lower the cost of earning attention. If anything, it made attention more scarce. We are drowning in things to watch and starving for reasons to watch them.
Even on a large scale, there is a reason the “Netflix effect” — which allows well-established shows to explode in viewership once they are available on the platform — is such a thing. Netflix built a flywheel by having the widest reach. It did not just make content available. It made discovery easy. That matters more than most people in tech want to admit.
This is the central problem for almost every new content tool. Just because you can generate the asset does not mean you can generate demand. In media, creation is the appetizer. Distribution is the meal. Note: it’s a shame HBO won’t benefit from Netflix’s flywheel, but I digress here.
Democratizing content creation is cool, but the reality is there is a big gap between mediocre and excellent storytellers.
Hollywood overpays for great storytellers for the same reason the Yankees overpay for left-handed pitching: scarcity. And narrative instinct is (very) scarce. The ability to understand what keeps a viewer emotionally locked in for 90 minutes, or eight episodes, is not just hard, it’s near alchemic.
There’s a reason some SFX wizards and content production companies flame out when handed a feature film to write and direct: they know how to render the explosion, but not the human being standing next to it. A lot of people in and around AI tech still underestimate that distinction.
The same logic applies here. AI can already make things look polished, surreal, cinematic, and technically impressive. That is something. But visual sophistication and emotional resonance are not interchangeable. One gets you a viral clip. The other gets you stickiness with a chance of revenues.
This means that most AI video content will continue to be incredibly impressive technically, but depressingly dull to watch. And dull is fatal in entertainment. Consumers do not reward effort, innovation or compute spend. They reward laughing, crying and generally caring about the characters on screen.
This is something I keep advising entertainment companies on: Stop thinking that personalizing the content is what people want. Intimacy does not equate relevance.
Executives love personalization because it sounds modern, measurable and vaguely inevitable. But the superpower of great content (and great brands, by the way) is not that it shows your little nephew as the hero, it’s that it becomes a shared experience that connects people — a cultural campfire, if you will.
The best entertainment is not about isolating us into smaller and smaller algorithmic bubbles. It gives us common grounds to react to. Something to argue about at dinner. Something to text a friend about immediately after it happens. A hit show is social currency. Its value is not only in how much one person enjoys it, but in how many people can enjoy it collectively.
What matters is that I can tell people they must watch The Pitt and we get to share how devastated we all felt when someone died. That is the point. The win is always: Did you see that last night?
There are categories where personalization has real value — education, wellness, shopping, even some forms of gaming. But mainstream entertainment is fundamentally a shared medium. We do not just consume stories to see ourselves reflected back. We consume them to feel plugged into something larger than ourselves.
OpenAI may be paying the price for trying to be everything, everywhere, all at once. Anthropic has proven to be more focused on what AI does best, while ChatGPT starts lagging there.
There is a general race to use AI in everything, but the reality is that AI is terrifyingly good at some things and expensively mediocre at others. Text, code, research, support, analysis: high frequency, high ROI. Video generation: expensive, compute-hungry, and still more demo than business.
That distinction matters. For a while, the market rewarded companies for looking ambitious. Now it is starting to reward them for being disciplined. We’re in a more mature phase where the winners will be the ones that understand where AI genuinely compounds value.
Knowing where to apply AI does matter, and this move is a necessary market correction now that AI needs to drive revenue fast, not just headlines and social media reactions.
Right now, when it comes to AI for entertainment, the gap between “impressive demo” and “durable business” is where hype goes to die.
OpenAI probably made the right call to shut down Sora.
Wall Street is weighing reports that the US has approached Iran with a plan to halt fighting.
Porter's career high is 50 points, from five years ago.
Making the jump straight from college head coach to big-league manager, Vitello represents a historic first for MLB and a much-needed infusion of energy for the Giants.
Patrick Mahomes tore his ACL on Dec. 14, putting his status for the 2026 opener in doubt.
Missed Jay-Z tickets when they first went on sale? You can still find resale seats.
The top official at the Transportation Security Administration offered new evidence that labyrinthine airport security lines and other travel woes are unlikely to recover quickly even when lawmakers reinstate funding.
Here's an early look at the top three quarterbacks in this class, including the Heisman Trophy winner, Alabama's rising signal-caller and a gunslinger who makes good decisions.
Silver's response comes after the National Basketball Players Association's statement about Cade Cunningham’s award eligibility status on Tuesday.
Zillow Group CEO shares some important perspective on the struggling US housing market.
