BrainChip recently announced its Akida neural networking processor. The processor uses chips inspired by the spiking nature of the human brain. It’s part of a growing effort to commercialize chips based on human neural structures. The new generation of chips could mean “more deep neural network processing capability in the future on portable devices, e.g., smartphones, digital companions, smartwatches, health monitoring, autonomous vehicles and drones,” Vishal Saxena, a professor of electrical and computer engineering at the University of Delaware told Lifewire in an email interview.
Brains on a Chip
BrainChip says the new boards could help usher in a new era of remote AI, also known as edge computing, due to their performance, security, and low power requirements. By mimicking brain processing, BrainChip uses a proprietary processing architecture called Akida, which is both scalable and flexible to address the requirements in edge devices. At the edge, sensor inputs are analyzed at the acquisition point rather than through transmission via the cloud to a data center. “I am excited that people will finally be able to enjoy a world where AI meets the Internet of Things,” said Sean Hehir, BrainChip CEO, in the news release. “We have been working on developing our Akida technology for more than a decade, and with the full commercial availability of our AKD1000, we are ready to fully execute on our vision. Other technologies are simply not capable of the autonomous, incremental learning at ultra-low power consumption that BrainChip’s solutions can provide.” Mercedes uses the BrainChip processor in its new Mercedes Vision EQXX concept car, promoted as “the most efficient Mercedes-Benz ever built.” The vehicle incorporates neuromorphic computing to help reduce power consumption and extend vehicle range. BrainChip’s Akida neuromorphic chip allows in-cabin keyword spotting instead of using power-hungry data transmission to process instructions. One significant advantage to chips designed like a brain, also called neuromorphic design, is potential power savings. Although researchers understand very little about the basis of cognition, a human brain only consumes around 20 watts of energy, Saxena said. “This is due to the fact that the brain performs ‘in memory computing’ and communication using spikes in an event-driven fashion, whereby energy is only consumed when a spike is emitted,” he added. Neuromorphic chips are a good fit for processor-intensive tasks like deep learning AI computers because they use much less power. The chips could also be helpful for edge devices like smartphones where battery power is limited, Saxena said.
Future Chip Brains
BrainChip is one of many start-ups focusing on brain-inspired chips, called neuromorphic design, including SynSense and GrAI Matter Labs. Intel is working on its Loihi neuromorphic chip, but it’s not yet available for purchase. The international research group IMEC in Belgium develops neural networks to develop better audio devices, radar, and cameras that react to certain events. Neural chips offer “the ability of on-line learning, making sensing systems adaptive to real-world variations (think of changing light conditions for cameras or variations person-to-person for wearables),” Ilja Ocket, a program manager at IMEC, told Lifewire in an email interview. Neuromorphic chips could also allow computers to see like humans. Prophesee is applying neuromorphic techniques to vision processing. The company’s approach is called event-based vision, which only captures and processes information that changes in a scene like humans do instead of a continuous stream of data for the entire locations that conventional cameras use. Neuromorphic chips could one day enable more intelligent sensors in devices like smart wearables, AR/VR headsets, personal robots, and robot taxis, Ocket said. The new chips could perform local AI tasks to learn from and adapt to local and changing environments. “All this without the need for cloud communication, hence enabling built-in privacy,” he added.
