PrismML, an innovative AI lab established by researchers from Caltech and guided by advisor Ion Stoica from UC Berkeley, has launched a compact version of its language models specifically optimized for smart glasses equipped with Qualcomm’s Snapdragon chips.
During the recent Snapdragon Summit, Qualcomm unveiled PrismML’s new 1-bit Bonsai LLM, designed to operate locally on AI smart glasses running on the Snapdragon AR1 Gen 1 Platform. This development reflects a significant advancement in portable AI technology.
As highlighted in a report by TechCrunch, one of PrismML’s major achievements lies in its ability to compress larger models, achieving a reduction of 4x without sacrificing performance on common benchmarks. The latest smart glasses model is a 2-billion-parameter version fine-tuned for tasks that combine vision and language. This allows users to inquire about their visual surroundings in real-time.
The overarching vision of PrismML emphasizes open-weight AI that maximizes the capabilities of existing devices, aiming to eliminate reliance on the privacy assurances of proprietary AI labs, which often demand significant computing resources.
While announcing a model compatible with Qualcomm’s technology is a notable milestone, details about any forthcoming smart glasses featuring PrismML remain undisclosed.
Editor’s Take
This development represents a pivotal move in making AI more accessible through everyday devices like smart glasses. By enabling a local processing model, users can benefit from real-time insights without compromising privacy. This could significantly expand the applications of AI in augmented reality, offering new opportunities for businesses and developers in the evolving tech landscape.
Source: techcrunch.com
