Sora is built on the technology behind OpenAI’s image generation DALL-E tool. It interprets user prompts, expands them into a more detailed set of instructions, and creates new videos using AI models trained on videos and images.
The quality of AI-generated images, audio, and video has improved rapidly over the past year, with companies like OpenAI, Google, Meta, and Stable Diffusion creating and selling more capable tools. We are racing to find a way. At the same time, democracy advocates and AI researchers warn that the tool is already being used to deceive and lie to voters.
This isn’t the first time such videos and audio have been created, with other companies building their own text-to-video AI generators. Google is testing something called Lumiere, Meta is developing a model called Emu, and AI startup Runway is already building a product to help filmmakers create AI videos. But AI experts and analysts said the length and quality of Sora’s videos exceeded anything seen before.
“We didn’t expect to see this level of consistent video generation for the next two to three years,” said Ted Underwood, a professor of information science at the University of Illinois. While he cautions that OpenAI likely selects videos that best show the model, “there seems to be a bit of a jump in capacity” compared to other text-to-video tools. said.
In Pakistan, former Prime Minister Imran Khan used AI to create a digital version of himself giving a speech, even though he is currently in prison. An ad supporting Florida Governor Ron DeSantis’ now-defunct campaign for the Republican presidential nomination used an AI voice generator to imitate former President Donald Trump’s voice.
The tech company building the tool says it monitors the use of its tools and has instituted several policies against its use to create political content. However, enforcement has been uneven. In January, OpenAI suspended a developer who created a bot for Democratic candidate Dean Phillips following a Washington Post report. Developers were creating similar bots for political candidates in the fall.
This rapid advancement in technology has people in a variety of industries, from film production to the news industry, scrambling to understand how technology will impact their work.
AI video generators are already making waves in Hollywood. Making a film is expensive, time-consuming, and requires the collaboration of dozens or hundreds of people. Some technologists theorize that AI could allow him to create a movie by himself with the same visual complexity as Marvel’s blockbusters.
“Look at how far we’ve come in just one year of image generation. Where will we be in a year?” Observer Michael Gracey believes that in the near future, AI tools like Sora will allow filmmakers to carefully control the output and create any type of video from scratch. I’m predicting.
“You don’t need a team of 100 or 200 artists over three years to create a feature-length animation,” he said. “That’s very exciting for me.”
At the same time, Gracie said a major problem is the fact that AI tools are trained without compensating the work of real artists. “It’s not nice to steal someone else’s creativity, work, ideas, and execution, and deny them fair recognition or financial reward.”
Mutare Nkonde, a visiting policy fellow at the Oxford Internet Institute, said the idea that anyone could easily convert text into video was interesting. But she also explores how these tools embed social bias, their impact on people’s lives, and how they transform hateful text and descriptions of tragic real-world events into harrowingly realistic footage. I’m concerned about my abilities.
Nconde said recent strikes by the Writers Guild and Actors Guild began to address issues surrounding the use of AI language tools in scriptwriting and the use of actor likenesses in AI-generated scenes. But tools like Sora raise new questions, she said, such as whether human extras are necessary. She asked, “From a policy perspective, do we need to start thinking about how to protect the human beings who should be involved with these tools?”
The quality of Sora’s videos, especially those meant to look like the real world, is higher than what most other AI companies have been able to produce to date.
Arvind Narayanan, a computer science professor at Princeton University, said Sora “seems to be significantly more advanced than any other video generation tool” based on the video OpenAI published Thursday. The result is likely to be “deepfake” videos that are harder for people to recognize as being generated by AI, he said.
If you look closely at some videos, you can still find many contradictions, he said. For example, in an X post, he pointed out that in a video of a Tokyo street scene, the positions of a woman’s right and left feet are swapped, and people in the background disappear after something passes in front of her.
Still, casual viewers may not notice such details, he added. “Sooner or later we have to adapt to the fact that realism is no longer an indicator of reliability.”
