How Edge AI Can Be a Driver for Green Technology

The world needs to go green fast, and edge AI technology can help––if it manages to leap a few hurdles.

Across various business, industrial, and household activities, opportunities abound for the reduction of emissions and waste. At this moment, fresh produce is spoiling away in the forgotten corner of a refrigerator. Meanwhile, a building is stacking up its energy bill as air-conditioning aggressively cools all the wrong corners.

For most people, daily life is all too rife with mundane concerns; conserving the planet requires a foresight that can be a lot to ask of the preoccupied mind. Thus, the proliferation of AI-enabled devices could go a long way in conserving the planet. In real time, they can monitor pollution, reduce water and energy waste, and much more––requiring minimal human exertion.

Nonetheless, four obstacles stand in the way of AI adoption: data insecurity, inaccessible costs, energy inefficiency, and logistical deployment difficulties. As an answer to these challenges, Kneron, a San Diego-based AI startup, has arrived first-to-market with lightweight algorithms, reconfigurable NPU chips, and the vision to build a democratized and private edge AI net.

Towards the edge AI net

Picture this: A consumer walks into Best Buy and checks out three edge AI nodes the size of tennis balls. One has a camera sensor, one has a chemical sensor, and one has a thermal sensor. All three of them have an NPU-enabled SoC. At home, the consumer installs AAA batteries in each node, links them to her home assistant, and puts them into her fridge. From an open-source app store, she then downloads a food waste reduction application which monitors and maintains the freshness of perishable vegetables through the nodes.

Three months later, she becomes too busy and no longer stocks up on fresh produce. She removes the edge AI node with camera sensor from her fridge and places it in her living room to reduce the air conditioning’s energy use by counting the number of persons in the room; she repurposes the edge AI node with chemical sensor and the edge AI node with thermal sensor for monitoring her garden compost. All she needed to do was to download new applications and unsubscribe from the one no longer in use––no new hardware is required, and no previous hardware is dumped into the trash.

Kneron’s vision of an edge AI net is an open software platform to market and deliver intelligent edge AI services to a private peer to peer mesh network of edge AI AIoT devices using cloud native technology. In delivering intelligent AI services, valuable data being generated are guaranteed data accuracy by AI and data integrity by employing blockchain technology.

The vision starts with Kneron’s ultra-low power high MAC efficient AI chip playing a crucial role at enabling edge AI computing to take off, especially mobile edge AI. The edge AI net technology takes it one step further enabling the proliferation of capable edge AI cameras and other sensors to communicate and collaborate with each other to create higher intelligence and value, while protecting user privacy with a peer to peer private mesh network. For example, location triangulation is an example of higher intelligence created by multiple AI cameras. The edge AI IoT devices can create intelligence through an array of various sensors and among diversity of sensors to create the power of sensor fusion.

This versatility of application is enabled by the fact that Kneron’s NPU microprocessors are reconfigurable, making them adaptable to any AI model. This means that the chip may be combined with a variety of sensors to deploy any possible application.

Green––but seamless and effortless

Founded in 2015, Kneron is an end-to-end edge AI solutions provider that serves as a single port of call for device manufacturers who want to integrate AI into their products. The company’s products include both NPU chips and AI models that can be incorporated into anything from autonomous cars to smart fridges, doorbells, or any Internet of Things device.

Attracted by these solutions, a list of prominent investors––including Horizons Ventures, Sequoia, Alibaba, and Qualcomm––have taken stakes in Kneron. Backed by robust funding and market-proven products, the company’s technologies are in a unique position to unlock the sustainability potential of edge AI by overcoming the challenges of data security, accessibility, energy efficiency, and ease of deployment.

 Data privacy and security issues have earned a foul reputation for AI and connectivity among some consumers, Kneron eliminates these concerns by creating a private edge AI net that operates independently of the cloud. And whereas AI-enabled products are available today only at high prices, Kneron’s solutions are cost-saving, making them accessible everywhere for everyone.

While achieving high inference performance, Kneron’s solutions are also inherently more lightweight and energy-efficient than those of most competitors. In the 2019 Face Recognition Vendor Test held by the National Institute of Standards and Technology, Kneron’s facial recognition algorithm––Kneron-003––scored the best aggregate rating among all lightweight models and outperformed several much bulkier models.

Enabled by this energy-efficiency, Kneron has collaborated with a partner to develop an AI-enabled door lock that can run on 8 AA batteries for up to 12 months. This energy-saving performance results from the company’s fundamental departure from mainstream industry goals. At present, the industry competes on producing the largest tera operations per second (TOPS.) Kneron, however, competes not just on operations per second, but operations per watt.

And yes, energy footprint is a real concern in democratizing AI. In the famous 40-day long experiment wherein AlphaGo defeated Lee Sedol, the machine employed 5,000 of Google’s proprietary machine-learning processor units and consumed 50,000 times more energy than did the human mind. Cloud-based AI requires large bandwidth and tends to produce higher energy footprints, both of which are tackled in the development of edge AI.

The energy-efficiency of Kneron’s solutions, alongside their flexibility, further help to overcome the logistical difficulties of installing AI––such as physical bulk, energy source, and update costs. The low energy requirement of Kneron’s solutions allow them to run simply on readily available batteries, eliminating the ridiculously cumbersome requirement to draw wires across a rice paddy or build a power outlet into a fridge.

Furthermore, the reconfigurability of Kneron’s chips and edge AI nodes make them durable long into the future. They adapt to new AI models and applications, reducing both electronic waste and the costs of deploying new AI applications.

Equipped with reconfigurable AI chips and award-winning algorithms, Kneron has been the choice solutions provider for several smart device manufacturers, including a leading global producer of air conditioners that optimize both user comfort and environmental sustainability. By democratizing AI from the cloud to edge, Kneron works to enable environmentally conscious practices that are both seamless and effortless.

The post How Edge AI Can Be a Driver for Green Technology appeared first on EE Times Asia.

from EE Times Asia https://ift.tt/3d6H25b