Silicon Labs on Monday showed off two new systems-on-chips and a new software toolkit that the vendor says will bring AI and machine learning to the Edge.
Chips include an artificial intelligence and machine learning accelerator, wireless radio, ultra-low-power capabilities, support for IP Matter and Zigbee-based connectivity protocol, and flash capability.
The toolkit enables developers to build and deploy AI and machine learning algorithms using TensorFlow.
As a supplier of silicon chips and software integrations that run on the IoT, Silicon Labs said the BG24 and MG24 chips support multiple wireless protocols and can be used for edge devices in home applications. intelligent, medical and industrial.
Solve some problems
Offerings like these aim to solve some of the problems enterprises face when bringing AI and ML to the edge, said Andy Thurai, analyst at Constellation Research.
Typically, AI/ML edge operations require many edge-optimized algorithms to run for inference, and they also need to be updated frequently.
Andy ThurayAnalyst, Constellation Research
“In general, AI/ML edge operations require many edge-optimized algorithms to run for inference, and they also need to be updated frequently,” Thurai said.
One of the reasons they are limited is power. Many peripheral devices use batteries to provide flexibility so that they do not need to be connected to a power source.
“If they’re not very low-power devices, those batteries have to be constantly changed, which can be very expensive to operate,” Thurai said.
Silicon Labs said its chips were very low-power, but Thurai said they had to be justified against other chips on the market.
Another problem with edge and IoT devices is that patchy Internet and wireless connectivity has had limited success in bringing AI and machine learning applications to the edge.
Having out-of-the-box wireless connectivity for sensors with Bluetooth and ZigBee can reduce costly implementation and integration processes, as sensors do not need to be connected to the cloud or a hybrid network , said Thurai.
Silicon Labs System-on-Chip
BG24 and MG24 are designed to perform difficult calculations quickly. According to Silicon Labs, machine learning calculations do not occur in the cloud, but rather on the local device, which speeds up decision making.
“This is true for anyone doing local edge processing,” Thurai said, adding that by processing data at the edge, users can avoid round trips to the cloud, latency and other issues.
“While many companies claim faster decision-making, the cloud round trip is typically in milliseconds, not seconds,” he continued. “For many applications, this is not an obstacle.”
Silicon Labs’ systems-on-chips (SoCs) also feature large flash and random-access memory capacities, according to Silicon Labs. The vendor said the chips can scale for multi-protocol support, Matter and machine learning algorithms trained for large data sets.
The systems also include Platform Security Architecture Level 3-Certified Secure Vault, a chip-based subsystem that provides security for IoT devices such as door locks, medical equipment, and sensitive deployments.
Secure Vault includes security features that address IoT threats and protect against hardware and software attacks.
“Security and privacy go hand in hand when it comes to smart homes,” said Bob O’Donnell, analyst at Techanalysis Research. “Having some support for that will also be important going forward.”
New hardware from Silicon Labs supports Matter, Zigbee, Bluetooth Low Energy and more.
A new approach
Adding AI to low-power battery devices is a new approach, O’Donnell said.
“We’ve heard about cutting-edge AI and things like that. But to be honest, it mostly requires big devices, like a smartphone,” O’Donnell said. “What these chips open up is the ability to do similar things conceptually, with battery-powered devices in the smart home and other types of applications.”
Since SoC chips need software to operate, O’Donnell said the accompanying software toolkit is important because many AI programmers are familiar with TensorFlow.
Besides TensorFlow, Silicon Labs has also partnered with vendors of AI and machine learning tools, including SensiML and Edge Impulse for the software toolkit.
Matter matters
Developers can use the SoCs as well as Silicon Labs’ Simplicity Studio with the Software Toolkit to build applications that communicate with each other using Matter.
Matter enables distinct IoT ecosystems to come together. For example, in a smart home scenario, Matter can help unify the device ecosystems of Apple, Google, and Amazon.
One challenge Silicon Labs may face is the slow evolution of Matter standards, which are often delayed in committees. This can further complicate already complex wireless connectivity tasks for edge and IoT devices, O’Donnell said.
“They’ve been doing this for a long time, so I think they’ve pretty much nailed it,” he said. “They’re combining the abilities they had in other chips in other chips, but that’s how this world works.”
“In theory, it should work out of the box,” he added.
More than 40 have started using the chips and software kit in a closed Alpha program, Silicon Labs said.
The chips and software toolkit will be available for mass deployment in April, the vendor said. Silicon Labs did not provide pricing information.
