The GPT-4o model introduces features like transparent backgrounds, beneficial for business users and creatives. Despite improvements, it still faces challenges like hallucinations and editing consistency. OpenAI promises rapid updates to address these issues. Ethical and legal concerns persist, with OpenAI asserting its model was trained on publicly available data and proprietary data from partnerships. Images generated with GPT-4o will include C2PA metadata to denote them as AI-generated, adhering to industry standards.
Key takeaways:
- OpenAI has integrated its 4o model into ChatGPT for native image generation, eliminating the need for separate use of Dall-E, though Dall-E remains available for preference.
- The new features, including the Sora AI video generator, are available for free and paid ChatGPT users, with enterprise and education users gaining access next week.
- The GPT-4o model enhances image realism and text legibility through reinforcement learning from human feedback, and introduces the ability to create transparent backgrounds.
- Despite improvements, the GPT-4o model still faces challenges such as hallucinations, editing consistency, and ethical concerns, with OpenAI promising rapid updates and using C2PA metadata for AI-generated images.