PrismML Introduces Small-Sized Language Models for Qualcomm-Powered Smart Glasses

PrismML, an AI startup founded by Caltech researchers, introduced a small-sized language model optimized for smart glasses powered by Snapdragon chips at the Qualcomm Snapdragon Summit.

Model Unveiled at the Qualcomm Summit

PrismML, founded by Caltech researchers and advised by UC Berkeley's Ion Stoica, announced AI models tailored for smart glasses. Chipmaker Qualcomm showcased this new technology during its Snapdragon Summit.

This specialized AI system developed can be run locally directly on smart glasses built on the Qualcomm Snapdragon AR1 Gen 1 Platform.

Features of the Bonsai LLM Model

The unveiled system relies on the 1-bit Bonsai language model, which shrinks large models by a factor of four while maintaining almost complete performance on standard benchmarks.

Specifically adapted for smart glasses, this version features two billion parameters and is equipped with visual and language processing capabilities. This enables users to ask real-time questions about the objects they are looking at.

Local AI Vision

PrismML's primary goal is to develop open-weight AI systems that run on-device and utilize available processing power more efficiently.

The startup presents this approach as an alternative to the privacy promises and ever-increasing processing power dependency of proprietary AI laboratories.

Future Plans and Market Release

Releasing a model for Qualcomm's chip is considered an important step toward realizing the startup's vision of on-device open AI.

However, no smart glasses model capable of running PrismML's software has been officially announced yet.