Sustainability of Edge Intelligence
EThe focus during the last decade on deep learning accuracy ignores economic, or environmental cost. Progress towards Green AI is hampered by lack of metrics that equally reward accuracy and cost and can help to improve all deep learning algorithms and platforms. We define recognition and training efficiency to compare deep learning in a universal fashion and compare it to published similar, but less universal approaches. It is based on energy consumption measurements for inference and deep learning, on recognition gradients, and on number of classes. Well-designed edge ASIC accelerators reach higher recognition and training efficiencies compared to cloud CPUs and GPUs due to reduced communication overhead. Sustainability of edge intelligence algorithms and platforms is assessed with deep learning lifecycle efficiency and life cycle recognition efficiency metrics that include the number of times models are used.
The overall impact of Green AI is compared with other sustainability measures in the information industry and artificial and natural intelligence efficiencies are compared leading to insights on deep learning scalability.
Bruno Michel received a Ph.D. degree in biochemistry and computer engineering from the University of Zurich and joined IBM Research to work on scanning probe microscopy and soft lithography. Later he improved thermal interfaces and miniatur¬ized convective cooling and demonstrated improved efficiency and energy re-use in sustainable datacenters, and photovoltaic thermal solar concentrators. He developed microfluidics, 3D packaging with interlayer cooling and electrochemical chip power supply to trigger a density roadmap to replace Moore’s law. Most recently he focusses on integration of IoT and wearable devices with efforts spanning from sensing principles over edge computing to multi- sensor data fusion and cognitive computing. Areas of particular interest are sustainable edge intelligence and Green AI. He is an IEEE Fellow as well as a member of the US National Academy of Engineering and the IBM Academy of Technology.
Wednesday, October 21, 2020
( ETHZ, Switzerland and Universita di Bologna, Italy)
From Near-Sensor to In-Sensor AI
Edge Artificial Intelligence is the new megatrend, as privacy concerns and networks bandwidth/latency bottlenecks prevent cloud offloading of sensor analytics functions in many application domains, from autonomous driving to advanced prosthetic. The next wave of "Extreme Edge AI" pushes signal processing and machine learning aggressively into sensors and actuators, opening major research and business development opportunities. In this talk I will focus on recent efforts in developing an AI-centric Extreme Edge computing platform based on open source, parallel ultra-low power (PULP) customized ISA-RISC-V processors coupled with domain-specific accelerators, and I will look ahead to the next step: namely three-dimensional integration of sensors, mixed-signal front-ends and AI-processing engines.
Luca Benini holds the chair of digital Circuits and systems at ETHZ and is Full Professor at the Universita di Bologna. He received a PhD from Stanford University. In 2009-2012 he served as chief architect in STmicroelectronics France. Dr. Benini's research interests are in energy-efficient computing systems design, from embedded to high-performance. He is also active in the design ultra-low power VLSI Circuits and smart sensing micro-systems. He has published more than 1000 peer-reviewed papers and five books. He a Fellow of the IEEE, of the ACM and a member of the Academia Europaea. He is the recipient of the 2016 IEEE Mac Van Valkenburg award and the 2020 ACM/IEEE A. Richard Newton Award.