IEEE/IEIE ICCE-Asia 2023
TODAY 2023. 12. 10
Advancements in AI for Mobile Consumer Electronics
: A Comprehensive Review
Prof. Seung-Wook Kim
- Division of Electronics and Communication Engineering of dept., Pukyong National University
In the era of mobile technology, the integration of AI into consumer electronics is driven by the need for personalized user experiences, efficient systems, and enhanced security. The speaker will present several state-of-the-art AI models and discuss how these models can be optimized for mobile platforms to improve performance despite hardware limitations. Designed for anyone from students to professionals interested in the intersection of AI and mobile technology, this tutorial provides a clear understanding of this rapidly evolving field.
Seung-Wook Kim received the B.S. and Ph.D. degrees in School of Electrical Engineering from Korea University, Seoul, in 2012 and 2019, respectively. In 2019, he was a Research Professor with the Semiconductor Research Center, Korea University. From 2020 to 2022, he was a Staff Researcher with the Samsung Advanced Institute of Technology, Suwon, South Korea. In 2022, he joined the Division of Electronic and Communication Engineering, Pukyong National University, Busan, South Korea, where he is currently an Assistant Professor. His research topics include computer vision, image processing, and machine learning.
Healthcare Sensors : Body Temperature and Blood Glucose Monitoring
Dr. Jeongseok Chae
- Principal research engineer of Dongwoon Anatech
Recently, we can monitor biological vital signs with cheap cost compared with medical grade equipment of hospital by using mobile devices such as smart watches.
Moreover, in this year, ChatGPT is a hot issue which shows tremendous achievement of artificial intelligence (AI) area. Healthcare service can be improved and commercialized more by using AI technologies and continuously sampled biological data. Gathering personal biological signs requires many sensor ICs.
In this tutorial, two biological sensor systems, body temperature and noninvasive ood glucose monitoring systems, will be shown in detail.
Jeongseok Chae received the B.S. and M.S. degrees in electronics engineering from Kookmin University, Seoul, Korea, in 2000 and 2002, respectively. He received the Ph.D. degree in electrical and computer engineering at Oregon State University, Corvallis, USA, in 2011. He was a design engineer with Samsung Electronics, Kiheung, Korea, from 2002 to 2006 working on battery protection ICs and display (TFT-LCD and AMOLED) driver ICs for mobile solutions. He was a member of technical staff with MaxLinear, Carlsbad, USA, from 2010 to 2011 working on high speed ADC design for DOCSIS modem application. He designed and/or developed analog blocks such as touch sensor, ambient light sensor, audio amplifier and galvanic isolator. Currently, he is a principal research engineer with Dongwoon Anatech, Seoul, Korea, working on healthcare SoCs. His research interests include high-resolution and low-power analog circuit and system design
Latency Processing Unit (LPUTM) for Acceleration of Large Language Model Inference
Prof. Joo-Young Kim
In this talk, I will present a new processor breed named LPU for accelerating the inference of hyperscale AI models. Optimized for transformer-based large language models such as OpenAI’s GPT and Meta’s LLaMA, LPU manages to execute an end-to-end inference with low latency and high throughput. LPU uses model parallelism and optimized dataflow that is model-and-hardware-aware for fast simultaneous workload execution among multiple devices. Its compute cores operate on custom instructions and support entire GPT operations including multi-head attentions, layer normalization, token embedding, and LM head. We implement the proposed hardware architecture on multiple AMD Alveo U55C datacenter accelerator cards and utilize all of the channels of the high bandwidth memory (HBM) and the maximum number of compute resources for high hardware efficiency. Finally, the multi-LPU server appliance achieves 1.49x speedup and 2.35× cost efficiency over the GPU supercomputer DGX A100 with state-of-the-art library (FasterTransformer), showing high performance scalability of 1.76x when doubling the number of devices. We are building LPU silicon IPs, compiler stack, and software development kit.
Joo-Young Kim received the B.S., M.S., and Ph. D degrees in electrical engineering from Korea Advanced Institute of Science and Technology (KAIST), Daejeon, South Korea, in 2005, 2007, and 2010, respectively. He is currently an Associate Professor with the School of Electrical Engineering, KAIST. He is also the Director of AI Semiconductor Systems Research Center. His research interests span various aspects of hardware design, including VLSI design, computer architecture, FPGA, domain-specific accelerators, hardware/software co-design, and agile hardware development. Before joining KAIST, he was a Hardware Engineering Leader at Microsoft Azure, Redmond, WA, USA, working on hardware acceleration for its hyper-scale big data analytics platform named Azure Data Lake. He was also one of the initial members of Catapult project at Microsoft Research, Redmond, where he deployed a fabric of FPGAs in datacenters to accelerate critical cloud services such as machine learning, data storage, and networking. He founded a AI silicon startup HyperAccel in Jan 2023 to build innovative AI processor/solutions for generative AI, making it sustainable for everyone.
Dr. Kim was a recipient of the 2016 IEEE Micro Top Picks Award, the 2014 IEEE Micro Top Picks Award, the 2010 DAC/ISSCC Student Design Contest Award, the 2008 DAC/ISSCC Student Design Contest Award, and the 2006 A-SSCC Student Design Contest Award. He also served as Associate Editor for the IEEE Transactions on Circuits and Systems I: Regular Papers (2020-2022), and is an IEEE SSCS Distinguished Lecturer for the term 2023-24.