AI Rolls Deep from the Factory to the Edge at CES
January 17, 2024
There were lots of jokes in Las Vegas last week about how CES is the world’s biggest auto show these days, (more on that in future stories), but it would be just as true to say the Consumer Electronics Show is the premier home of the leading commercial AI use cases anywhere.
We saw Generative AI and Machine Learning engines right next to statistical analysis on big data sets, personal assistants, advanced chatbots, and numerous interesting Edge AI Implementations.
Let’s take a look at a very few of the ones we saw in every hall and nearly every booth.
Giant of Industrial and Enterprise technology Siemens announced a brought spectrum of news from the show, but one of note here is the partnership with Microsoft to bring Generative AI to bear on industrial edge intelligent systems. Taking the form of a Gen AI-powered digital assistant, Siemens Industrial Copilot is designed to enhance human-machine collaboration and boost productivity in manufacturing, the partners said.
Copilot will allow users to generate, optimize, and debug complex automation code, Siemens asserted, and therefore shorten simulation times from weeks to minutes. The automation and process simulation information will come from Siemens’ open digital business platform, Siemens Xcelerator, and will be augmented with the Azure OpenAI Service. An early adopter will be automotive supplier Schaeffler AG.
The Intelligent Edge
Aizip is a company focused on tinyML and Edge AI with a stated goal of being computer intelligence into ubiquity. Working with scientists from MIT, UC Berkeley, UC Davis, and UC San Diego, the company has created an automated AI-design pipeline called Aizipline that uses AI models to develop other AI models like a self-evolving system. Aizip says it can be used for data generation, model design, and model testing for starters, thanks to its AI design automation technology.
At this year’s show, we saw live demos of the tiny AI models at work in both consumer and industrial applications. There was a running motor with a vibration sensor and loaded with the company’s AI IC chip that enables preventative maintenance that in the demo detected the smallest inefficiency in operation and issued an alert to check it and then cleared the alert automatically when the problem was removed. Aizip says this edge AI can be yours starting at 10 cents per IC, US.
Morse Micro, the Wi-Fi HaLow silicon company, dropped a slew of intelligent wireless systems during the week, and showcased several demos that illustrated the potential of Wi-Fi HaLow technology, a Wi-Fi standard specifically designed to support the IoT market.
The booth was full of lots of applications from the Smart Home to the Factory, but the Edge AI-powered security cameras linked with HaLow were a high point. The low-power, high-performance IEEE 802.11ah compliant SoCs and modules loaded in the devices provide 10x the range, 100x the coverage area and 1000x the volume of traditional Wi-Fi solutions, Morse says.
This can enable industrial automation, building management systems, smart meters, and innumerable other Edge AI use cases. Most notably, the company announced a new HaLow router with partner Edgecore Networks, the Edgecore EAP112.
The companies said this router is ideally suited to processing at the AI-enabled Edge because of its robust design and multi-protocol connectivity, including Wi-Fi 6, Wi-Fi HaLow, 4G LTE, BLE, Zigbee, and Thread. With all these options and the interconnected networks, data processed at the edge can be sent from any environment to central processing in the cloud or on-premise servers.
With a real forward-looking vision, Orbbec is layering deep learning ML into its computer vision solutions with the newly announced Gemini 2 XL, which is a wide FOV RGB-D PoE camera designed for AI-based vision developers. This long-range, 3D camera delivers accurate and reliable data in various lighting conditions from pitch black to outdoor daylight, the company said. The Orbbec SDK enables easy setup, and the RGB-D quality can provide data suitable for AI-driven visual recognition and tracking systems that need to detect small objects even in highly reflective environments, the company says. Moreover, the included deep learning algorithms can generate a phase map around small objects and fixtures, overcoming shadow, speckle pattern size, illumination condition, and other challenges.
This is of course only a taste of the AI applications we saw at the show, and we’ll be digging into even more over the next few weeks, but it’s safe to say that 2024 might just shape up to be the year that AI really did finally take its rightful place.