Despite being a prototype with some bugs, R2X demonstrates Nvidia's ambition to merge generative video game technology with advanced LLMs, aiming to create a human-like AI assistant. The avatar can assist with tasks like using Adobe Photoshop’s generative fill feature and processing documents through a local retrieval augmented generation feature. Nvidia plans to open-source these avatars in 2025, allowing developers to integrate their preferred AI software. The company is also working on enabling R2X to join Microsoft Teams meetings and potentially take actions on desktops, though these features are still in development.
Key takeaways:
- Nvidia unveiled a prototype AI avatar called R2X at CES 2025, designed to assist users on their PC's desktop by navigating apps and processing files.
- R2X can be run on popular large language models like OpenAI's GPT-4o or xAI's Grok, and it can interact with users through text and voice.
- The avatar can view and assist with applications on the screen, but it has faced issues like incorrect instructions and screen viewing failures during demos.
- Nvidia plans to open-source these avatars in the first half of 2025, allowing developers to integrate their preferred AI software products or run the avatars locally.