Home / Technology / Tiny AI for Smart Glasses Unveiled
Tiny AI for Smart Glasses Unveiled
25 Sep
Summary
- AI Lab PrismML created tiny language models for smart glasses.
- Models run locally on Qualcomm Snapdragon chips for real-time use.
- The goal is open-weight AI on devices, reducing compute dependency.
AI Lab PrismML, with roots in Caltech and guidance from UC Berkeley, has engineered a specialized version of its compact language models for smart glasses. These models are designed to run locally on Qualcomm's Snapdragon chips. At Qualcomm's recent Snapdragon Summit, the 1-bit Bonsai LLM was showcased, highlighting its capability to operate directly on smart glasses equipped with the Snapdragon AR1 Gen 1 Platform.
PrismML is recognized for its ability to significantly reduce the size of larger AI models, achieving a 4x compression while maintaining high performance on benchmarks. The smart glasses variant is a 2-billion-parameter model, specifically tuned for vision and language tasks, enabling users to query their surroundings in real-time. The startup's broader objective is to advance open-weight AI that functions on end-user devices, maximizing existing computational resources.
This development represents a key step towards PrismML's ambition of providing an alternative to proprietary AI services, which often rely on extensive cloud computing and raise privacy concerns. While no specific smart glasses models featuring PrismML technology have been announced yet, this collaboration with Qualcomm marks progress toward their vision of on-device, accessible AI.
