Brainchip has worked out a way of converting traditional algorithms to State Space Model (SSM) algorithms resulting in way smaller chips.
See Edge AI and Vision Alliance description by Tony Lewis concerning his presentation quoted below.
The implications for a much smaller chip are huge. Think small devices including mobile phones.
For example, the possibility of Pico or a slightly larger version of Pico in mobile phones for example for 'game changing cybersecurity ' is huge.
Also, the possibility for other small devices is limitless.
This has largely gone under the radar.
"Explore the future of AI at the edge withM Anthony Lewis, CTO ofBrainChip, at the Embedded Vision Summit. He'll contrast state-space models (SSMs) with transformers, showing how SSMs, with their read-only architecture, reduce power consumption and chip area. Discover how SSMs achieve real-time interactivity and how new techniques enable distillation-based migration from transformer models like Llama without losing performance."
See Edge AI and Vision Alliance description by Tony Lewis concerning his presentation quoted below.
The implications for a much smaller chip are huge. Think small devices including mobile phones.
For example, the possibility of Pico or a slightly larger version of Pico in mobile phones for example for 'game changing cybersecurity ' is huge.
Also, the possibility for other small devices is limitless.
This has largely gone under the radar.
"Explore the future of AI at the edge withM Anthony Lewis, CTO ofBrainChip, at the Embedded Vision Summit. He'll contrast state-space models (SSMs) with transformers, showing how SSMs, with their read-only architecture, reduce power consumption and chip area. Discover how SSMs achieve real-time interactivity and how new techniques enable distillation-based migration from transformer models like Llama without losing performance."