Podcast: How a Low-Power Edge AI Chip Company is Driving Intelligence in Consumer Devices

As home security solutions such as security cameras are now used in several households, there is an increasing need for AI (Artificial Intelligence) inferencing at the edge. Typically, the security solution relies on machine learning models to identify objects or faces in the CCTV footage – for example differentiating between a cat and a human. The data is then sent to the cloud for analysis and sent back to the device. But this can be problematic when there is poor internet connectivity.

Ergo, a tiny 7x7mm Edge AI chip from a company called Perceive, aims to solve issues with AI inferencing at the edge. The chip enables rapid processing on edge devices, for example facial recognition, or alerting to certain sounds, such as glass breaking or a dog barking. This can trigger actions without resorting to cloud-based systems. This type of solution can also offer enhanced data security and user privacy, as the data does not leave the device. The edge AI inference chips can be used in connected devices such as smart speakers as well, where many commands can be processed on the device, rather referring to the cloud. There can be many other applications in the future including drones, autonomous vehicles, and much more.

In the latest episode of ‘The Counterpoint Podcast’, host Peter Richardson is joined by David McIntyre, VP of Marketing at Perceive. David talks about AI inferencing at the edge using a tiny chip called Ergo. He deep dives into problems solved by inferencing on edge devices over the cloud, use cases, and savings made related to space onboard, costs and power. The podcast discussion also focuses on potential applications where solutions like Perceive’s Ergo chip can be used.

Hit the Play Button to Listen to the Podcast

You can download the podcast transcript here.

Podcast Chapter Markers

00:58 – A little bit about David, his role at Perceive, and what solutions the company offers.

02:31 – What is an edge inferencing device and what problems does it solve?

06:01 – How do you go about training the model for inferencing at the edge?

10:27 – The Ergo chip and its headline features?

13:17 – The number of sensors that can be used in Ergo chip-based devices?

14:09 – Does the solution need any external memory?

15:32 – Privacy and security aspects when keeping inference data locally?

18:21 – Where Ergo is being deployed?

20:38 – The support Perceive offers to device makers?

22:31 – What are you most excited to see with edge AI inferencing applications in the coming years?

Related Posts