At Computex 2024, we introduced Project G-Assist, a tech demo highlighting how AI can enhance the PC experience for gamers and creators. We are now releasing an experimental version of this system assistant for GeForce RTX desktop users through the NVIDIA app, with support for laptops to follow. Modern PCs, while more powerful, have become increasingly complex to manage, with users facing countless hardware and software configurations to optimize performance across various components. Project G-Assist aims to streamline this process by functioning as an AI assistant that operates locally on GeForce RTX AI PCs. It can optimize games and system settings, track performance statistics, and control peripheral settings—all through simple voice or text commands. Unlike cloud-based AI models requiring online access and subscriptions, G-Assist utilizes the GPU's power directly, making it responsive, free, and capable of operating offline. Utilizing a Llama-based Instruct model with 8 billion parameters, G-Assist provides advanced language processing in a compact form, keeping up with advancements in AI research. Activating G-Assist is as simple as pressing Alt+G, allowing users to optimize graphics settings or check GPU temperatures. During this process, users might experience a slight dip in performance, but the GPU swiftly returns to full power afterward. G-Assist requires specific PC components and is built with NVIDIA ACE technology, which is also used by game developers for AI NPCs.

Companies like MSI, Logitech, and HP are already incorporating ACE to create custom AI assistants. NVIDIA encourages community development for G-Assist, promoting expansion through a GitHub repository with sample plugins and instructions. Existing plugins include Spotify for music control and Google Gemini for enhanced interactions. Users can generate their plugins using a ChatGPT-based “Plugin Builder, ” facilitating easy integration with G-Assist. NVIDIA plans to open G-Assist to the AI community, enabling integration into various workflows and applications. Users can download G-Assist from the NVIDIA app, with ongoing updates expected to enhance its capabilities. Feedback from users is vital for future developments, and suggestions can be submitted through the app.