by The maker of ChatGPT is now diving into the world of videos created by artificial intelligence (AI).
Meet Sora – a new OpenAI text-to-video generator. The tool, unveiled by the San Francisco-based company on Thursday, uses generative AI to instantly create short videos based on written commands.
Sora is not the first person to demonstrate this type of technology. But industry analysts point to the high video quality of the tool on display so far, and note that its introduction is a significant step for both OpenAI and the future of text-to-video generation.
Still, as with everything in the rapidly growing AI space today, such technology raises concerns about possible ethical and societal implications. Here’s what you need to know.
What can Sora do and can I still use it?
Sora is a text-to-video generator – creating videos up to 60 seconds long based on written prompts using generative AI. The model can also generate a video from an existing still image.
Generative AI is a branch of AI that can create something new. Examples include chatbots, such as OpenAI’s ChatGPT, and image generators such as DALL-E and Midjourney.
Getting an AI system to generate video is newer and more challenging but relies on some of the same technology.
Sora isn’t yet available for public use (OpenAI says it’s in contact with policymakers and artists before officially releasing the tool) and there’s still a lot we don’t know. But since Thursday’s announcement, the company has shared a handful of Sora-generated video samples to show off what it can do.
OpenAI CEO Sam Altman X has also adopted the platform formerly known as Twitter to invite social media users to submit prompt ideas.
He later shared realistically detailed videos that responded to clues such as “two golden retrievers podcasting on top of a mountain” and “a bike race on the sea with different animals as athletes riding bikes with a drone camera view”.
Although videos generated with Sora can depict complex, highly detailed scenes, OpenAI notes that there are still some weaknesses – including some spatial and cause-and-effect features.
For example, OpenAI adds on its website, “a person might bite into a cookie, but afterwards, the cookie might not have a bite mark”.
What other AI-generated video tools are out there?
Sora OpenAI is not the first of its kind. Google, Meta, and the Runway ML startup are among companies that have demonstrated similar technology.
Still, industry analysts emphasize the apparent quality and considerable length of Sora’s videos shared so far.
Fred Havemeyer, head of US AI and software research at Macquarie, said the launch of Sora was a major step forward for the industry.
“Not only can you make longer videos, I understand up to 60 seconds, but also the videos that are being created look more normal and seem to respect physics and real life more,” said Havemeyer.
“You’re not getting as many ‘uncanny valley’ videos or clips on the video feeds that look … unnatural”.
While there have been “tremendous advances” in AI-generated video over the past year – including the introduction of Stable Video Diffusion last November – Forrester senior analyst Rowan Curran said there was a need for such videos ” stitching together” for consistency of character and scene.
The consistency and length of Sora’s videos, however, represent “new opportunities for creatives to incorporate elements of AI-generated video into more traditional content, and now even generate full-blown narrative videos from one or a few cues,” Curran added. the Associated Press via email on Friday.
What are the potential risks?
While Sora’s capabilities have wowed viewers since Thursday’s launch, concerns remain about the ethical and societal implications of using AI-generated video.
Havemeyer points to the significant risks in the potentially very narrow 2024 election cycle, for example.
He introduced a “magical” way to generate videos that could look and feel realistic, which presented a number of questions within politics and beyond, he said. – targeting fraud, propaganda and misinformation concerns.
“The negative externalities of generational AI will be a critical topic of debate in 2024,” said Havemeyer. “It is a substantial issue that every business and every person must face this year”.
Tech companies are still calling the shots on controlling AI and its risks as governments around the world work to catch up.
In December, the European Union agreed on the world’s first comprehensive AI rules, but the action won’t take effect until two years after final approval.
On Thursday, OpenAI said it was taking important safety steps before making Sora widely available.
“We are working with red teams – domain experts in areas such as misinformation, hate content, and bias – who will be testing the model adversarially,” the company wrote.
“We are also building tools to help detect misleading content, such as a detection classifier that tells when Sora generated a video”.
OpenAI Vice President of Global Affairs Anna Makanju reiterated this while speaking on Friday at the Munich Security Conference, where OpenAI and 19 other technology companies pledged to work together voluntarily to combat AI-generated election deepfakes.
She noted that the company was releasing Sora “in a very careful way”.
At the same time, OpenAI revealed limited information about how Sora was built.
OpenAI’s technical report did not reveal what image and video sources were used to train Sora — and the company did not immediately respond to a request for additional comment on Friday.
Sora’s release also arrives in the midst of the lawsuits against OpenAI and its business partner Microsoft by several authors and The New York Times regarding its use of copyrighted works of writing to train ChatGPT.
