Topline
ChatGPT maker OpenAI stepped up the race in generative artificial intelligence Thursday when it unveiled its text-to-video generation tool, Sora, viewed as an impressive but potentially dangerous step in the booming AI economy amid concerns about disinformation spread.
Key Facts
Sora, developed by Microsoft-backed OpenAI, which is also behind text and image generators ChatGPT and Dall-E, is being tested for safety and AI experts say it is a significant development on existing tools.
“Game on,” responded Cristóbal Valenzuela, co-founder and CEO of Runway, an AI video company that launched a Gen-2 update to its AI video model just months ago and is already available for anyone to use.
Stability AI, another leader in AI image generation, unveiled generative video model Stable Video Diffusion in November, but said it was “exclusively for research” and not ready for “real-world or commercial applications at this stage” (there is a waitlist, however, though the company hasn’t said when it will be released).
Pika Labs, a new but competitive player in AI video, launched its generative AI video tool, Pika 1.0, in November and in late December announced “the wait(list) is over” and made it available to everyone.
Google, which recently released its next-generation AI model Gemini 1.5, revealed research and demo clips from its Lumiere model, though it’s unclear when it plans to make the tool publicly available and it noted the need for robust tools to combat bias and malicious content to ensure its technology can be deployed safely.
Meta, which boasts a stable of apps including Facebook, Instagram and WhatsApp, first launched a video generation tool in 2022, called Make-A-Video, which produced noticeably artificial, distorted and blurry content but was nevertheless a significant achievement at the time.
Meta teased a new text-to-video model, Emu Video, late last year and said almost all people using it ranked it above Make-A-Video in terms of quality and its faithfulness to the text prompt, and though it did not give an indication of when the tool will be released, CEO Mark Zuckerberg said Emu will be integrated into Facebook and Instagram.
Elon Musk’s AI startup xAI, which he hopes will rival industry leaders like DeepMind and OpenAI, has not announced plans for a text-to-video generator, though it’s possible this capability could be folded into its “rebellious” AI chatbot, Grok, in the future.
Amazon has previewed both a generative AI chatbot for workers, Q, and an AI image generator, though the e-commerce giant has not disclosed when it plans to roll the tools out or whether more advanced video generators are on the way.
Can I Use Sora Yet?
No. Sora isn’t available to the public and OpenAI has not indicated a timeline for when it will be. The company said it will be “taking several important safety steps ahead of making Sora available in OpenAI’s products,” adding that it is working with experts in areas including misinformation, hateful content and bias to test the model.
Key Background
OpenAI’s concern over how its video generation tools can be misused is echoed by practically every firm working in the field. The ability to craft realistic and convincing content can help spread dangerous misinformation or damaging material like fake pornographic content featuring apparently real people, abusive content and inflammatory announcements from politicians and companies. In the past, many of the biggest social media companies like Facebook and Twitter, now X, have been accused of not doing enough to tackle the huge amounts of misinformation and hate speech on their platforms. And with technological advances, experts warn the risks have only grown. The risks are especially acute during an already challenging election year and firms are under more scrutiny than ever to stamp out misinformation. The advances to video generators will make it easier to fabricate false video content, make that content more believable and also make it harder to detect. Fake scenes and images have already emerged of Hillary Clinton (who was implausibly endorsing Ron DeSantis), President Joe Biden (announcing a military draft) and former president Donald Trump (getting arrested and being chased by police).
Crucial Quote
“Ur a wizard sama,” Stability founder and CEO Emad Mostaque said in a post on X. The tech leader was responding to a video posted by Altman (sama is the OpenAI CEO’s X handle) of a wizard-themed video generated by Sora.
Further Viewing
While not available to the public, Altman invited people to suggest text prompts to show off Sora’s text-to-video capabilities. Videos produced include various sea creatures like (legged) dolphins and turtles riding bikes on the ocean surface, an instructional gnocchi video hosted by a grandmother in a Tuscan kitchen, a futuristic drone race on Mars and several duck-dragon hybrids carrying an adventuring hamster through the sky.
