How did we build the Next Generation of AI PCs - Copilot+ PCs?
summary
Windows now ships 40+ models that run locally on an NPU, crossing the ~2 billion‑parameter threshold so devices host larger LLMs like Phi Silica and run image creator, OCR, translations and studio effects concurrently. Recall builds a local vector database from continual OCR and encodings, all executing with NPU efficiency under 4 watts so a demo runs all models on the device with GPU idle and still leaves capacity to play a game.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 1:54NPU technology
- 3:33AI performance and efficiency
- 5:38Enabling new experiences
- 7:17Performance demonstration
- 9:01Developer tools and closing
related talks
30mCloud? No Thanks! I’m Gonna Run GenAI on My AI PC
45mPrivacy-First In-Browser Generative AI Web Apps
13mAccelerating AI algorithms
40mFueling the AI Revolution with Gaming
36mApplied AI on the edge
42mAI and Self-Driving Cars
18mSoftware is the New Fuel, AI the New Horsepower - Pioneering New Paths
52mPhysical AI: Crafting Resilient Systems with Emotional Intelligence