A New Smart Frontier: AI-driven Decision Making at the Edge
To increase the value of services, organizations are demanding greater capabilities at the point of service or care. This usually consists of a node at the network edge. Many decisions can be made from the massive volumes of data generated daily. Artificial intelligence can serve as a smart assistant that enables organizations to make faster better decisions without human intervention.
Interview with Mr. Steen Graham, General Manager of Edge.AI Scale, Intel Corporation
According to Gartner, around 10% of enterprise generated data is created and processed outside traditional centralized data centers or clouds. Conversely, this figure is expected to reach 75% by 2025. Edge AI allows enterprises to capture large amounts of data and act upon its insights close to the source. AI and deep learning tap the expertise of data scientists and combine it with knowledge from all data sources to efficiently react or make forecasts in real time as data is being acquired. The current COVID-19 pandemic has accelerated the adoption of AI applications across sectors such as retail, industry, health, and smart cities. To meet the fast-growing demand for edge AI applications, Intel® offers a wide range of processors and development tools to speed up the process from conceptual development to prototyping, stress testing, and pilot runs.
Powerful tools to fast-track edge Al application development
Mr. Steen Graham, General Manager of Edge. AI Scale at Intel Corporation, pointed out that AI is becoming embedded in the fabric of computing and pervasive across all industries. For example, healthcare leverages AI for imaging-based diagnosis. This allows GE Healthcare to offer radiologists intelligent pneumothorax detection at the point-of care in just a few seconds. By leveraging advances in AI, healthcare providers can enhance medical image analytics in ways that may help improve accuracy in diagnoses and imaging processing time. Industrial manufacturing uses robotics for defect detection, equipment monitoring, and real-time inventory assessments. Cities and transit points adopt AI to enforce face mask regulations and conduct body temperature monitoring. Safer shopping experiences are made possible by natural language processors (NLP), which enable contactless vending and payment, as well as spatial distancing.
Mr. Graham elaborated on their offers to developers. The range of AI applications and deployment locations requires a broad array of choice regarding performance, power, and total cost of ownership (TCO) envelopes. Intel’s technologies include processors, accelerators, and programmable devices, such as the Gen 3 Intel® Movidius™ Vision Processing Unit (VPU)—codenamed Keem Bay—which is purpose built for edge AI and computer vision. It boasts more than 10x the deep learning inference performance of the Movidius™ Myriad™ X VPU while consuming comparable amounts of power. Introduced in November 2019, the newest generation VPU achieves great performance and is a power efficiency achievement. Mr. Graham pointed out that it delivers noticeable performance-per-watt among AI accelerators such as GPUs. Efficiency is becoming an important benchmark for power-constrained applications at the network edge. The VPU architecture was designed to offer performance-per-watt efficiency advantage via minimized data movement on the chip. (Readers can find more information at https://intel.com/vpu).
To help develop algorithms, the Intel® Distribution of Open Visual Inference and Neural Network Optimization (OpenVINO™) toolkit allows developers to quickly build solutions that emulate human vision and scale across multiple architectures simultaneously such as VPU, CPU, and GPUs. OpenVINO™ toolkit also enables performance optimization to reduce the inference time of computer vision models. For instance, an existing Advantech platform with an available PCIe slot and an Advantech VEGA card powered by Intel® Movidius™ Myriad™ X VPU can be used to accelerate AI and utilize the host CPU processor, providing scalable performance for edge applications requiring multiple cameras.
To enable developers to design their solution faster and move quickly from prototype to production, Intel® DevCloud for the Edge offers full access to hardware platforms hosted in Intel’s cloud environment designed specifically for deep learning. Developers can deploy and test AI model performance using the Intel® Distribution of OpenVINO™ toolkit, as well as across Atom®, Core™, Xeon®, and VPU technologies, allowing them to optimize before they buy.
Ecosystem that leverages expertise and bridges technologies
Building a complete ecosystem means incorporating compelling hardware, software, and solutions purpose built and optimized for industry-specific use. Mr. Graham added that “with decades of partnership with Advantech and others, Intel has built an unparalleled ecosystem of specific-use edge hardware.” Collaboration across the Intel ecosystem brings together a range of expertise and abilities, not only speeding up the development of edge AI solutions but also shortening time to concrete business outcomes. Intel also boasts over 300 market ready solutions, such as autonomous mobile robots to manage inventory, cold chain monitoring solutions.
Advantech and Intel are working closely to deploy the latest edge AI technology on three fronts. First, Advantech and Intel collaborate through the early access program on Atom®, Core™, Xeon®, and Movidius VPUs to offer a diverse range of form factors that meet industry specific needs. Second, close collaboration enables developers using solutions like the OpenVINO™ toolkit and WISE-PaaS platform to get the best performance from Intel’s hardware solutions. Third, Intel showcases industry-specific solutions driving business outcomes for end users through many initiatives including Intel’s market-ready solutions; Advantech has a comprehensive portfolio of products and solutions ranging from intelligent retail to equipment monitoring and machine vision, and much more.
Together, Advantech and Intel offer a robust set of edge AI technologies that enable industrial transformation, help provide important healthcare solutions, and deliver accelerated business outcomes.