From TechCrunch · Julie Bort · · 1m 30s
PrismML's tiny AI arrives on Qualcomm smart glasses
Caltech-founded startup brings compressed language models to Snapdragon-powered wearables, offering on-device AI without relying on cloud services.
In brief
Caltech-founded startup brings compressed language models to Snapdragon-powered wearables, offering on-device AI without relying on cloud services. PrismML demonstrated its compressed language models on Qualcomm smart glasses at Snapdragon Summit, shrinking AI models by four times while keeping performance intact…
Tiny AI for smart glasses
PrismML, founded by Caltech researchers and advised by UC Berkeley's Ion Stoica, has created compressed language models for smart glasses running Qualcomm's Snapdragon chips.
Snapdragon Summit showcase
On Wednesday at Qualcomm's Snapdragon Summit, the chipmaker showcased PrismML's one-bit Bonsai LLM running locally on AI smart glasses built on the Snapdragon AR one Gen one Platform.
Shrinking models by four times
PrismML shrinks larger models substantially by four times while retaining almost all of their performance on standard benchmarks. The smart glasses version is a two-billion-parameter model tuned for vision and language.
Real-time vision questions
Wearers can ask what they are looking at in real time. The model runs locally on the device, enabling instant responses without sending data to cloud servers.
Open-weight AI vision
PrismML's larger goal is open-weight AI that runs on devices and makes better use of computing power they already have. The startup pitches this as an alternative to depending on proprietary AI labs.
No glasses announced yet
Releasing a model for Qualcomm's chip is a step toward that vision. But no smart glasses running PrismML have been announced yet.
In short
PrismML demonstrated its compressed language models on Qualcomm smart glasses at Snapdragon Summit, shrinking AI models by four times while keeping performance intact.
