AI Tech / news
PrismML shrinks its 1-bit Bonsai LLM for Qualcomm smart glasses
PrismML, an AI lab founded by Caltech researchers, has built a version of its tiny language models for smart glasses running on Qualcomm's Snapdragon chips. Qualcomm showcased the 1-bit Bonsai LLM at its Snapdragon Summit on Wednesday.
The model is a 2-billion-parameter system tuned for both vision and language, allowing a wearer to ask in real time what they are looking at. It runs locally on AI smart glasses built on Qualcomm's Snapdragon AR1 Gen 1 Platform, rather than in the cloud.
PrismML's core claim is compression: it shrinks larger models substantially, in this case by a factor of four, while retaining almost all of their performance on standard benchmarks. The lab was founded by researchers from Caltech and is advised by UC Berkeley's Ion Stoica.
Qualcomm put the 1-bit Bonsai LLM on display Wednesday at its Snapdragon Summit, the chipmaker's annual hardware event. The demonstration points to on-device processing as a route for AI features in wearable hardware.
The startup frames its work as a step toward open-weight AI that runs on devices and makes better use of the computing power they already contain. It pitches that approach as an alternative to relying on the privacy promises of proprietary AI labs and their need for ever more compute.
No smart glasses running PrismML have been announced so far.