CNN
—
OpenAI, a prominent player in artificial intelligence, has unveiled its latest AI model named Sora, touted for its ability to generate “realistic” and “imaginative” 60-second videos based on concise text prompts.
According to a recent blog post, Sora can craft minute-long videos from textual cues, incorporating various elements such as multiple characters, specific movements, and intricate background settings.
The blog post highlighted that the model not only comprehends the user’s prompts but also interprets how these elements manifest in the physical realm.
OpenAI’s primary objective is to refine the AI model further to assist individuals in tackling real-world challenges through interactive solutions.
Sora represents the company’s newest endeavor following the success of ChatGPT, a widely recognized chatbot, propelling advancements in generative AI technology. While “multi-modal models” and text-to-video capabilities are not groundbreaking concepts, OpenAI distinguishes Sora with its purported length and precision, as noted by Reece Hayden, a senior analyst at ABI Research.
Hayden emphasized the potential impact of such AI models on digital entertainment sectors, envisioning a surge in personalized content distribution across various platforms, particularly in television for enhancing narrative experiences.
Despite its promising features, OpenAI acknowledged Sora’s ongoing development phase, acknowledging certain “weaknesses” such as spatial inaccuracies, like confusing left and right orientations, and challenges in depicting cause and effect relationships. For instance, creating a video of someone biting a cookie might not accurately show the bite mark immediately after.
OpenAI’s current focus revolves around ensuring the safety and reliability of Sora. The company plans to collaborate with experts to scrutinize the model, specifically addressing concerns related to misinformation, offensive content, and bias. Additionally, OpenAI is devising tools to identify and combat misleading information effectively.
Initially, Sora will be accessible to cybersecurity experts, known as “red teamers,” for risk assessment purposes. Moreover, visual artists, designers, and filmmakers will have the opportunity to explore Sora’s capabilities and provide valuable feedback on its potential applications in creative industries.
This recent development coincides with OpenAI’s continuous enhancements to ChatGPT. One notable feature being tested allows users to manage ChatGPT’s memory, enabling personalized conversations by retaining past interactions or clearing previous dialogues for privacy.
