Sora is the most hyped bot since ChatGPT

For more than two years, every new AI announcement has been in the shadow of ChatGPT. No company model has eclipsed or equaled that initial fever. But perhaps the closest a company has come to this excitement was when OpenAI debuted its video-generating AI model Sora last February. Enticing clips — woolly mammoths kicking up snow clouds or Pixar-like animations of adorable fluffy animals — promised a breathtaking future in which anyone could create high-quality clips by typing simple text prompts into a computer program.

But Sora, which wasn’t immediately available to the public, remained just that: a teaser. The pressure on OpenAI has increased. In recent months, several other major technology companies, including Meta, Google and Amazon, have unveiled their own video generation models. Today OpenAI finally responded. “This is a launch we’ve been looking forward to for a long time,” the startup’s CEO Sam Altman said in an announcement video. “We will launch Sora, our video product.”

In the announcement, the company said that paying ChatGPT subscribers in the US and several other countries can use Sora to create their own videos. Unlike other tech companies’ video generation models, which remain in preview versions or are available exclusively through enterprise cloud platforms, Sora is the first video generation product that a major technology company puts directly into users’ hands. Chatbots and image generators like OpenAI’s DALL-E have already made it possible for anyone to effortlessly create and share detailed content in just seconds – threatening entire industries and driving profound changes in online communication. Now the era of video-generating AI models will only make these changes more profound, rapid and bizarre.

The key word from OpenAI this afternoon was product. The company is billing Sora not as a research breakthrough but as a consumer experience — part of the company’s ongoing commercial lull. When founded in 2015, OpenAI was a nonprofit organization with a mission to build digital intelligence “to benefit humanity as a whole, without being limited by the need for financial return.” Today it sells products and deals like any other tech company looking for revenue. OpenAI added a for-profit arm in 2019 and is reportedly considering removing control of its nonprofit board entirely in September. Sora’s marketing even represents a shift from February, when OpenAI introduced the video generation model as a step toward the company’s lofty mission of creating technology smarter than humans. Bill Peebles, one of Sora’s lead researchers, told me in May that video would open up “some avenues to AGI,” or artificial general intelligence, by allowing the company’s programs to simulate physics and even human thought. To create a video of a soccer game, Sora may need to model both the aerodynamics and the psychology of the players.

Today’s announcement, meanwhile, was preceded by a review from Marques Brownlee, a YouTuber known for his reviews of gadgets like iPhones and virtual reality headsets. Altman wore a hoodie emblazoned with the word Sora. Altman and the Sora product team spoke for more than 17 minutes; Peebles and another researcher spoke for one minute and 45 seconds, mostly praising how the company is launching a “turbo” version of Sora that is “much faster and cheaper” to bring a “new product experience” to the to bring to market.

The Sora release comes on the third of the “12 Days of OpenAI,” a period in which a new product is released or demonstrated to users every day. What the company has announced is certainly more like a product than a breakthrough in computer science: an elegant interface for creating and editing videos with features like “Remix,” “Loop,” and “Blend.” So far, many of Sora’s works have been impressive and even miraculous. The company hasn’t built a new, smarter bot so much as a user interface in the style of iMovie and Premiere Pro.

Videos created with Sora by OpenAI employees and early access users are already making their way onto social media, and a flood of users around the world will follow. For more than two years, low-cost and easy-to-use generative AI models have turned anyone into a potential illustrator. Soon everyone could become an animator too. This poses an obvious threat to human illustrators and animators, many of whom have long raised the alarm about generative AI taking away their livelihoods. Sora and similar programs also raise the specter of disinformation campaigns. (Sora videos come with a visual watermark, but OpenAI’s highest subscription tier, which costs $200 per month, lets customers create clips without one.)

But job displacement and disinformation may not be the most immediate or significant consequences of OpenAI’s third day. Both happened without Sora, although the program speeds up each problem: production studios were already experimenting with AI products for companies to create videos, such as a recent Coca-Cola Christmas commercial. And cheap, lower-technology methods of creating and spreading false information have been wildly successful in their own right.

What the mass adoption of video-generating AI products could significantly change is the way people express themselves online. Over the past year, AI-generated memes, cartoons, caricatures and other images, sometimes called “slop,” have flooded the internet. This content, much of it clearly generated by AI and not intended to deceive—a medium of crude self-expression, not sophisticated deception—may have been technology’s biggest impact on the 2024 presidential election. The fact that anyone can produce such images provides the opportunity to immediately express undecided feelings about an undeveloped world through an immediately digestible image. As my colleague Charlie Warzel has written, such content is intended to be consumed “quickly and with little or no thought about the initial reaction of the limbic system.”

A flood of AI-generated videos could provide even more powerful ways to visually communicate confusion, charged emotions, or persuasive propaganda—perhaps a much more lifelike version of the recent, low-quality AI-generated video of Donald Trump and Jill Biden in a fistfight, for example. Sora could take over TikTok and similar short-video platforms, just as AI image-generating models have distorted Facebook and changed the way people on X show their support for political candidates.

Sora’s takeover of the web is not guaranteed. Back in May, Tim Brooks, another Sora researcher who has since joined Google, compared the current state of the program to GPT-1, the earliest version of the programs underlying ChatGPT, which is currently in its fourth generation. OpenAI repeated the analogy today. This comparison has failed as the company has become more and more profit-oriented: GPT-1 was very preliminary research, a concept before a proof of concept, and four years away from the release of ChatGPT. Sora may still be untapped as a path for AGI, but it has become a full-fledged product almost ten months after OpenAI announced the model. Such early-stage technology may not represent significant progress toward curing cancer, solving the climate crisis, or other ways the startup claims AI could benefit all of humanity. But it could be all OpenAI needs to boost its bottom line.