PrismML Develops LLM for Qualcomm-Powered Smart Glasses
PrismML has developed new small language models for Qualcomm-powered smart glasses. Details of the 1-bit Bonsai model, which runs on Snapdragon processors, are in our news.
PrismML, founded by Caltech researchers and advised by figures leading in the field of artificial intelligence, announced that it has developed a new language model optimized for smart glasses running on Qualcomm’s Snapdragon platform. Introduced at the Snapdragon Summit last Wednesday, this technology is particularly noteworthy for its 1-bit Bonsai LLM model, which can run natively on wearable devices with the Snapdragon AR1 Gen 1 platform.
With this innovative step, PrismML aims to prove that large language models (LLMs) can operate efficiently despite hardware limitations. This 2-billion-parameter model, which allows users to make sense of their surroundings in real time, offers both visual and language processing capabilities.
- Developed by PrismML, the 1-bit Bonsai model can run locally on the device.
- The new model is designed to be compatible with the Snapdragon AR1 Gen 1 platform.
- The model manages to maintain its performance while reducing its size fourfold compared to standard models.
- Glasses users can receive real-time information about the objects they are looking at.
PrismML Aims to Increase Efficiency with Small Models
PrismML’s core thesis in the technology world is to significantly miniaturize existing artificial intelligence models without sacrificing performance. The company makes large-scale models four times smaller, enabling them to run on devices with limited processing power, such as smart glasses. This approach not only saves on hardware costs but also reduces concerns about privacy. Processing information on-device instead of transferring it to cloud systems offers a significant advantage in terms of protecting personal information.
The Open Source AI Era Begins in the Industry
The company’s primary goal is to develop open-weight AI models that maximize the processing power of devices. Positioned as an alternative to the closed systems offered by private AI laboratories, this technology allows developers to have control over their own hardware.
By combining this vision with a strong hardware partner like Qualcomm, PrismML aims to set a new standard in the wearable technology market.
No Device Yet for Mass Production
Although this new model, running on Qualcomm’s Snapdragon platform, is revolutionary in terms of technology, a commercial smart glasses model incorporating this software has not yet been released.
PrismML’s move serves as a roadmap for hardware manufacturers and provides valuable insights into how future smart glasses designs will be shaped by AI integration. It is expected that these small but effective models will be featured in more devices in the coming periods.
Do you think that AI models running locally on the device could completely replace classic cloud-based systems in terms of closure and performance? You can share your thoughts and expectations on this matter in the comments section.
