18 Aug AI at the Extreme Edge – MicroAI AtomML™
It is estimated that there are somewhere between 10 and 12 billion Internet of Things (IoT) devices active worldwide today. That number is expected to grow to at least 30 billion by the year 2030.
IoT and Industrial IoT (IIoT) devices and machines are now omnipresent. From a few exotic applications just a couple of decades ago, these devices now play a significant role in our personal and professional lives. From smart consumer devices to industrial and manufacturing equipment, IoT and IIoT devices and machines are now everywhere.
With that massive proliferation of devices and machines has come significant advancements in the artificial intelligence (AI) and machine learning (ML) technologies that make those assets more intelligent. A primary driver for those advancements has been the need to move AL and ML closer to the IoT device or machine.
The Challenge – Maximizing AI’s Potential
As the number of IoT-enabled devices and machines began to explode it became apparent that moving embedded intelligence as close to the asset as possible would create several operational advantages. There were several questions: How close is close? How do we get there? What about devices with low CPU capacity? What are the cost and ROI implications?
It was clear that a paradigm shift in embedded AI delivery would be required to meet the challenges faced by virtually every industry segment.
Intelligence at the Extreme Edge – MicroAI AtomML™
How close is close? What would be the next step in the evolution of Edge AI? MicroAI AtomML™ has answered those questions by bringing AI and ML to the extreme edge. MicroAI™ Atom provides developers with the means to embed AI onto the microcontroller (MCU) of a device or machine. Developers are now empowered to embed and train AI models onto assets that have limited CPU and memory capacities. This has opened the door to significant operational advantages when compared to traditional edge AI solutions. Advantages that include the following:
Asset/MCU centricity. AI can be embedded and trained onto individual assets. This brings big infrastructure intelligence down to the asset MCU level. A semi-supervised learning engine that trains AI models on any Cortex M Class MCU.
Hardware reduction. Up to this point, AL and ML solutions required investment in costly extraneous hardware to support the AI/ML delivery ecosystem. By embedding and training intelligence on the device MCU, adopters can minimize or eliminate investments in support hardware.
Less cloud dependence. Traditional Edge AI solutions require heavy investment into cloud support services. All AI-generated data must be transmitted to a cloud environment for processing, analysis, and storage. This results in higher AI system cost, increased latency, and greater security risks.
Greater flexibility. MicroAI™ Atom brings big infrastructure data in a footprint that is more flexible and more scalable. Asset-centric companies can now deploy AI gradually (starting with a single asset), fine-tune their processes, and then quickly scale to include a larger set of assets.
Faster response. By its very nature, AI at the extreme edge delivers data and insights faster than legacy AI solutions. Asset data is processed locally–right at the asset—eliminating the need to transfer data to the cloud. This provides asset operators and stakeholders with quicker insights into asset performance and enhances the ability to collect, and act upon, predictive maintenance data.
Enhanced Security. By embedding and training in the local environment, MicroAI™ Atom eliminates the need to ship asset data to the cloud. It allows equipment to develop models at their deployed locations and eliminates the cyber-risk associated with exposure of sensitive data to a cloud environment.
Reduced cost. Historically, cost has been one of the primary barriers to AI and ML implementation. By moving intelligence to the extreme edge, MicroAI™ Atom helps eliminate that cost barrier by reducing/eliminating hardware costs, minimizing cloud costs, and improving implementation efficiency.
AI at the extreme edge is no longer a futuristic vision; it is now a reality. Adopters of this cutting-edge advancement will now have the AI-fueled technology to reach new levels of operational efficiency and market leadership.
