Kuwait News Agency analyzes videos generated by OpenAI
Sora: AI revolutionary text-to-video tool
OpenAI has finally unveiled Sora, an artificial intelligence tool that transforms text instructions into video— another step in the AI realm. The launch of Sora, which means sky in Japanese, created an uproar amongst the media industry and social media in the US, while the Washington Post commented that this move would pave the way for producing high-quality video clips from scratch. Kuwait News Agency (KUNA) analyzed videos generated by OpenAI and showed high-definition (HD) minute-long clips, only produced by inserting texts.
OpenAI, backed by Microsoft, said Sora could generate videos from still photos in addition to adding visual content. It said Sora was in a testing phase to assess critical areas for harm or risk but was grateful to have access to a number of visual artists, designers, and filmmakers to give feedback to help develop the model to be most helpful for creative professionals. OpenAI CEO Sam Altman urged users outside OpenAI to present proposals on how to produce videos and to give the public an idea of the capabilities of AI. OpenAI invented the ChatGPT, GPT-4, and Dall-E.
It took a further step and offered Sora, giving the company another leap towards AI. Although AI-generated videos were already produced by tech giants Meta and Google in the past year, videos produced by Sora in terms of quality, accuracy, and length were at another level and
generated within a minute of the request. Sora produced a clip on the following text: a girl walking in a Tokyo street filled with colorful and warm lights and moving signs. The girl wears a black leather jacket, covering a long red dress and wearing black shoes. The girl carries a black purse, wears sunglasses, wears lipstick, and walks in a confident way on a wet street, reflecting lights from above. Many people walk in this street.
One of the complicated descriptions that Sora produced was a scene of the gold frenzy in California, US, in 1848, as well as a clip of an old lady blowing candles and archeologists discovering a plastic chair in the desert and taking extra care in cleaning it. Another challenging text was: Beautiful, snowy Tokyo city is bustling. The camera moves through the bustling city street, following several people enjoying the beautiful snowy weather and shopping at nearby stalls. Gorgeous sakura petals are flying through the wind along with snowflakes.
The Tokyo scene showed how Sora believed that these scenes were real and depicted accurate camera movement. Sora is another breakthrough in AI, revolving around algorithms processing huge quantities of data through network engineering, enabling machines to break AI barriers to go beyond self-learning to thinking, imagining, analyzing, visualizing, and listening to voices and generating reactions. Rehab Ismail, a multimedia professor at the American University in Cairo (AUC), predicted Sora would make quick steps to digital media and replace existing programs used to produce videos, documentaries, and videographics.
“Journalists in newsrooms always look for new ways to produce videos and attract the public. And they already started producing AI photos and now we have videos ... the future is open to all possibilities,” she told KUNA in an interview. “We need lots of experiments to learn how to finesse the new skill and use it in newsrooms. In a short time this technology will be the normal even for the regular user,” she said. “The world is changing remarkably fast which adds additional responsibilities on journalists to develop their skills in order to absorb this change,” added Ismail.
Media organizations, she added, were worried about this expansion of technology, particularly regarding “intellectual property,” one of the pressing questions about the spread of AI. Ismail said there “are no real laws organizing intellectual property rights when we talk about the content of AI products. “Training AI to use the work of others without compensating them is a big problem that needs intensive research by academic and media institutions,” she added.
There are other challenges facing the public as well including misinformation and disinformation, facing a dilemma on what to believe. Ahmad Esmat, a professor of information technology at AUC, voiced concern that Sora would contribute to spread of “deepfake,” especially during elections and conflicts. Speaking to KUNA, Esmat, a specialist in verifying digital information, warned media institutions because “flood of AI” lacked proper investments aimed at verifying the contents.
Experts said humanity has only seen a fraction of what AI could do. “I did not expect to see this level of ongoing video production in two or three years from now,” Ted Underwood, an information technology professor at the University of Illinios, told the Washington Post, reacting to the capabilities of Sora. A new study, prepared by tens of researchers in workshops hosted by Chatham House, expected AI to dominate news production and digital content by 2026.
The study did not recommend a confrontation with AI but using it to make media more efficient, amidst regional and international conflicts affecting media jobs worldwide. The NBC News Channel expected Sora to affect around 25 percent of the video and movie industries in the US in 2026. Observers said it was hard to predict the influence of AI on the future of media but anticipate that this year will be decisive in the spread of AI applications, which a few years ago were only fantasies.