OpenAI hasn't set a firm timeline for releasing its impressive Sora video AI model. The company is currently in discussions with policymakers and seeking partnerships within the entertainment industry.
According to a statement to CNBC, OpenAI's Sora model remains in the research phase "due to conversations with policymakers." The company first unveiled Sora publicly in mid-February.
Bloomberg reported in late February that OpenAI wanted to avoid safety risks, especially in light of this year's U.S. elections. This caution may be related to the aforementioned ongoing discussions with policymakers. The model also reportedly faced technical challenges at the time.
In early March, OpenAI officially stated that there was still no release schedule for Sora. However, OpenAI CTO Mira Murati later announced that Sora would "definitely" be released this year. She noted that the generation costs need to be reduced first, as the system is currently "much, much more expensive" than existing AI systems.
Potential legal issues related to training on YouTube data could also be a factor. YouTube CEO Neal Mohan has warned OpenAI about this possibility, though OpenAI hasn't commented on whether Sora was trained using YouTube content. OpenAI CTO Mira Murati famously said in an interview that "she doesn't know."
OpenAI courts Hollywood creatives
Ahead of Sora's release, OpenAI is reaching out to Hollywood. The company is meeting with film studios, media executives and talent agencies in Los Angeles to build partnerships and encourage filmmakers to incorporate Sora into their work. Actor Ashton Kutcher is among those testing the model and expects it to have a far-reaching impact on the film industry.
OpenAI is also giving artists access to unreleased AI tools, including Sora, as part of an exhibition at New York's Strada Gallery. The "Strada Nuova: New Road" exhibition runs for three weeks and features works by diverse artists, including researchers, academics, and creatives working at the intersection of physical and digital art.
Since Sora's debut in February, the video AI market has developed rapidly. Several Chinese companies, including KLING, Vidu, Jimeng AI, and MiniMax, have introduced their own models. Runway ML has also unveiled two new AI models that represent significant advances in quality and efficiency, though they don't yet match the level of OpenAI demonstrated with Sora. Midjourney is preparing the release of its video model.
Sora's developers compared the model to GPT-1, the precursor to modern language models, viewing it as proof that generative AI models for video are scalable. Further scaling could lead to emergent capabilities, enabling groundbreaking new applications. It's worth noting that the commercialization of GPT models only began in earnest with GPT-3.
