NEWS  AI  March 14, 2024  Reading time: ~1 minute)

Max (RS editor)

A month ago OpenAI unveiled its latest marvel: Sora, a cutting-edge text-to-video generative AI model. Initially showcased to a limited audience of red teamers, Sora's imminent public launch is poised to democratize and revolutionize the realm of video creation and editing.

In a recent interview with The Wall Street Journal, Mira Murati, OpenAI's Chief Technology Officer (CTO), offered insights into the company's ambitious plans for Sora.

Murati revealed that the eagerly anticipated AI tool is slated for release "this year", with potential availability in mere months. Moreover, OpenAI aims to empower users with the capability to manipulate and refine the videos generated by Sora, signaling a significant step forward in AI-driven content creation.

However, the path to realizing Sora's full potential comes with its challenges. Murati disclosed that powering Sora incurs substantial costs, surpassing those associated with other AI models. Despite this financial hurdle, OpenAI is committed to making Sora accessible at comparable rates to DALL-E, its renowned text-to-image AI counterpart.

Like its predecessors, Sora relies on extensive data training to deliver meaningful outputs. Murati, when questioned about the specifics of Sora's training data, remained guarded, citing the utilization of publicly available or licensed datasets.

While acknowledging partnerships with platforms like Shutterstock, Murati remained uncertain about Sora's reliance on content from major social media platforms.

Amid concerns over potential misuse and dissemination of disinformation, OpenAI has implemented safeguards, including watermarks on generated videos and limitations on depicting public figures. However, the efficacy of these measures in thwarting misinformation remains uncertain, given the historical vulnerability of watermarks to circumvention.

 IMAGES CREDITS: OPENAI