A video circulated on X that showed a picturesque Tokyo from above with snowflakes falling and people bustling along the sidewalk. However, the imagery didn’t come from a photographer, drone or anywhere someone has physically been. The video was completely AI-generated by Sora, a new text-to-video model created by OpenAI.
Thursday, Feb. 15, OpenAI announced its new text-to-video model by displaying several videos the technology had created. The public is not yet able to use the technology, but OpenAI CEO Sam Altman told ABC News that he is first working with artists, policymakers, and others before releasing the new tool to the public.
Sora is not the first AI-generated imagery but some contend it is the most refined looking version of the text-to-video technology. In April, The New York Times reported on Runway AI’s version of text-to-video being released. Runway AI’s text-to-video ads featured choppy videos showing people and animals in fantastical settings.
However, Sora is hard to distinguish from real life. If a user didn’t know the images were created by artificial intelligence, they may assume they were real people or events.
While AI enthusiasts may be quick to brainstorm ideas for fun, experts warn the advancements in text-to-video AI may have the potential to be used for nefarious purposes as well.
According to NBC News, the Federal Trade Commission released proposed rules Thursday that would make it illegal to create AI impressions of real people, which includes protections for people in government and business.
“This action in the light of surging complaints around impersonation fraud, as well as public outcry about the harms caused to consumers and to impersonated individuals,” the agency wrote. “Emerging technology, including AI-generated deepfakes, threaten to turbocharge this scourge, and the FTC is committed to using all of its tool to detect, deter, and halt impersonation fraud.”
OpenAI told NBC News that it is working with experts to test Sora, so it is not used to create misinformation, hateful, or bias.
OpenAI has not disclosed what imagery and video sources are being used to train Sora and has offered little information about how it was built, according to euronews. The New York Times is reportedly suing OpenAI over its use of copyrighted works to train ChatGPT.
Hollywood writers have also expressed concerns about AI technology. According to Axios, In September, the Hollywood writers’ strike ended after an agreement was reached with studios on AI restrictions to protect writers’ products and credits. As part of the agreement, studios must also disclose if any material given to writers has been generated by AI or incorporate AI-generated material.
While some may be concerned about a potential Pandora’s Box unleashed by the power of AI, Altman did allow for some fan collaboration on Thursday. Altman used prompts by X users for the Sora model to create content based on people’s suggestions.
“A wizard wearing a pointed hat and a robe with white stars casting a spell that shoots lightning from his hand and holding an old tome in his other hand,” said one X user.
Another X user requested, “An instructional cooking session for homemade gnocchi hosted by a grandmother social media influencer set in a rustic Tuscan country kitchen with cinematic lighting.”
Altman then displayed the videos exactly as requested, simply by typing the request into the text-to-video model known as Sora.
Related Stories