By Floria Cui Yi
On February 15, 2024, OpenAI announced a new AI model named Sora, which can create videos up to 60 seconds in length from text instructions or text mixed with an image.
Sora now can produce videos up to one minute in length while adhering to the user’s request and preserving visual quality. (Photo: Getty Images)
According to OpenAI, Sora can create complex scenes with several actors, distinct motion styles, and precise background and subject details. The model’s linguistic comprehension allows it to produce engrossing characters vividly conveying emotions. It can also generate a video with multiple shots that maintain the same visual style and characters.
Currently, Sora is only accessible to red teamers (cybersecurity professionals who assess the harms and risks), as well as a few visual artists, designers, and filmmakers. It’s not yet available to the public. Nonetheless, there are debates worldwide regarding whether the future development and use of Sora will lead to crisis.
David Alexander Palmer, a sociology and anthropology professor at the University of Hong Kong, argues that the development of Sora can bring about a series of social crises in a sociological context. “Sora, together with the VR headset released by Meta last week, combines the experience of AI, VR, and physical worlds. People will become further alienated from their bodies and the direct experience of the physical world as they get addicted to the digital world. Detached from the physical world, they are unable to connect their bodies to the natural environment, and they are unable to experience profound joy and meaning in life.”
He also pointed out potential crises at the level of social relations, as the mediation of social relationships through VR AI will open the possibility for psychological manipulation and control by tech companies and governments to an unimaginable degree.
Sora’s ability to create stunning videos also pressures the arts and creative industries. A 3D Artist and YouTube influencer, Grant Abbit, expressed his concerns in a recently posted YouTube video.
“It’s a definite threat that Sora will put people from the art industry out of work. I have heard particularly freelance concept artists who are not getting as much work. It’s difficult to say for definite as the industry has gone through quite a slowdown recently, but it’s hard to deny that this is an obvious consequence of this kind of technology. ”
— Grant Abbit
For amateurs making money on digital drawing and animation production, the emergence of Sora also brings a potential crisis to them. Olivia Au, a 20-year-old university student, sells her digital drawings and animations on social media platforms for pocket money, is worried about the future. “I only make short animations but it’s still time-consuming. Sora is efficient and cost-effective in making short-duration videos. If everyone can use Sora, no one will buy my works.”
This is an animation Olivia Au posted on Instagram on June 29, 2023. The 12-second work took her a week.
Many believe that Sora needs to be more competent to produce an industry or social crisis.
Dirk Schnieders is a senior lecturer at the University of Hong Kong who focuses on computer vision and artificial intelligence. He thinks that Sora’s performance, while remarkable, is only in line with expectations given the rapid growth of the textual descriptions field and the considerable computing resources OpenAI has access to. Other companies have also been investing in adding a time dimension to 2D imagery for quite some time. And the methodologies behind Sora are detailed in published research. He assumes that Sora will not create a technology monopoly or disrupt the AI industry environment.
Sora has obvious limitations in spatial and cause-and-effect elements. The character’s expression is always strange, the hands are always unnatural, and the control of the number of moving objects isn’t stable. It may not be apparent in a very short video, but the more people watch the more they realize it’s not accurate.
Hongjun Wang, a PhD student studying large multimodal model at the University of Hong Kong, has the same feeling.
“I do not view Sora as a true breakthrough. Rather than creating a technological revolution alone, I see Sora continuing the gradual progress made across the AI safety field. Additional research will be needed to address challenges like robustness, transparency and avoid unintended behaviors. Its success also depends on how readily other researchers can adopt or improve its techniques. Overall, Sora moves the needle but does not inherently shake up the industry and continued work is still needed.”
— Hongjun Wang
Some believe that Sora will not bring about a social crisis or change as long as people are more sober, address it, manage the negative consequences, and create policies around it. Ioana Sendroiu, an assistant professor at the University of Hong Kong, researches how people navigate social changes. From the perspective of problem-solving challenges, she argues that the fear of a potential crisis can be brought on by the public’s fear of the unknown. “People also panicked when ChatGPT came out, but now it doesn’t seem as capable as people thought it would be. Ultimately, Sora is what it is, and we shouldn’t demonize it.” She said.
Reported by Floria Cui Yi
Edited by Catherine, Rita