Abstract: BrainChip Akida: The Power-Efficient Brain for Edge AI
Session details:
In the era of the Internet of Things (IoT), the demand for intelligent, on-device processing is growing exponentially. However, this intelligence must be delivered with minimal power consumption, a challenge for traditional AI accelerators. BrainChip's Akida™ neuromorphic processor addresses this challenge by pioneering a new approach to AI. Unlike conventional processors that consume significant power by processing all data, Akida is inspired by the human brain, utilizing a unique event-based processing methodology. This means it processes only essential sensor inputs, activating only when it detects meaningful information. This paradigm, based on spiking neural networks (SNNs), allows for continuous, real-time AI processing while operating at microwatt to milliwatt power levels.
This ultra-low power consumption is transformative for the sensors market. Sensors for vision, audio, and other data streams can now be equipped with on-chip AI for applications in industrial automation, medical monitoring, and smart infrastructure. This capability enables devices to operate for years on a single coin cell battery, reducing maintenance costs and unlocking new use cases. Furthermore, by performing inference and on-device learning directly on the chip, Akida reduces latency, improves data security, and minimizes reliance on cloud connectivity. This approach allows for quicker, more reliable, and private intelligent sensing at the very edge of the network.
This presentation by BrainChip CEO Sean Hehir, will delve into several use cases with commercial advantages of the Akida platform, demonstrating how this is driving the future of the sensors in wearables, always on remote sensing and battery powered sensor markets by delivering high-performance, ultra-low-power, and secure AI solutions for a new generation of intelligent devices. To support this impact, several cases studies that demonstrate this will be shown, including
Akida in a battery-powered wearable medical device from Onsor Technologies that predicts epileptic seizures one hour prior to onset, creating a better life for those that suffer fro epilepsy;
Akida in satellite-based imaging systems for remote sensing applications in space with an integration by FrontGrade Gaisler;
Akida in drone-based device for classifying swimmers and surfer in remote beaches from Arquimea
