In a recent video, Sam Altman teased the upcoming release of OpenAI’s new model, rumored to be called Orion (GPT-5). His cryptic tweet mentioned his love for the Midwest’s night sky and winter constellations, which many interpreted as a hint towards the Orion model, a prominent winter constellation. This follows a similar pattern from his previous hints about the release of the Strawberry model, suggesting that we might expect Orion’s debut before the end of the year.
The video also covered Microsoft’s revamp of its Co-Pilot product, which aims to enhance AI integration across its suite of applications. The new features include Co-Pilot Pages, which allows users to create collaborative documents using AI, and Co-Pilot in Excel, which can generate Python code and create charts based on natural language prompts. Additionally, Co-Pilot in PowerPoint can generate entire presentations from simple requests, showcasing a significant leap in AI’s ability to assist in productivity tools.
Meta’s AI developments were also discussed, particularly regarding the training of its LLaMA model on publicly available Facebook and Instagram data. Despite Mark Zuckerberg’s previous statements about not using user data for training, it was revealed that all public posts since 2007 have been utilized for AI training, raising concerns about user privacy and data usage. This highlights Meta’s strategy of leveraging its vast data resources to differentiate itself in the AI landscape.
The video introduced several new AI projects, including Pixol, a vision model from MRAW AI that allows users to ask questions about images, and E11, a new prompt engineering library designed to simplify the process of working with language models. Adobe also announced its Firefly video model, which generates videos from text prompts, emphasizing its commitment to safe and high-quality AI tools. These innovations reflect the ongoing evolution of AI technologies and their applications across various industries.
Lastly, the video touched on the launch of World Labs by the “Godmother of AI,” Fei-Fei Li, which aims to develop AI with 3D perception capabilities. This initiative seeks to advance beyond traditional generative models by creating spatially intelligent AI that can understand and reason about the world in three dimensions. The video concluded with a discussion on the future of video games, highlighting Tencent’s GameGenO, a model designed for generating open-world video games, showcasing the potential for AI to revolutionize gaming experiences.
