In an exclusive interview with The Wall Street Journal, Mira Murati, Chief Technology Officer of OpenAI, revealed new details about Sora, the company’s innovative AI text-to-video generator recently announced. The interview covered a variety of different topics, including the type of content that Sora is capable of producing and the security measures implemented to combat misinformation.
Murati emphasized that Sora will have multiple layers of security to prevent misuse of this AI, especially regarding its potential influence on elections taking place worldwide. As a notable aspect, Sora will refuse to create “images of public figures”, such as the president of the United States.
To ensure the authenticity of the generated content, watermarks will be added to the videos. These will include a transparent OpenAI logo in the bottom right corner and possibly the source of the content will be added as another indicator of authenticity, using metadata to provide information about the origin of the digital media.
Looking towards the future, the developers of Sora have ambitious plans. According to Murati, it is planned to include sound in the generated videos to increase their realism. In addition, editing tools are being developed that will allow video creators to correct errors made by the AI.
And the thing is, despite its advanced technology, Sora is not exempt from errors, something that was evidenced when requesting a video of a robot stealing a woman’s camera and, instead, the woman partially transformed into a robot. Murati acknowledged that there is room for improvement and stated that, although AI is “quite continuous, it is not perfect”.
Finally, in the interview, the issue of nudity was also addressed. Murati indicated that OpenAI is collaborating with artists to determine what kind of nudity will be allowed with AI. The team seems willing to allow “artistic” nudity, although deepfake non-consensual fakes will be prohibited.
