Abstract
AI has been around for 60 years, and it has experienced several booms and winters. The AI winters were caused by the then fragile nature of AI and the complexity of real-world scenarios. The current AI boom is mainly motivated by Large Models (language, visual, multimodal, etc.) that have much better cross-task generalization ability compared with the previous approaches. However, for AI to truly land in the real world, we need AI to be scalable from one scenario to another, enable it to deal with corner cases and bring business value that is more than the investment. If we look at some of today’s AI forerunners, the answers to the above questions are still unclear. We all know that it takes a long time to turn an innovation into a product. In this talk, we will discuss how to land AI in the real world, from an industry perspective and from a landing perspective.
In the real world, there are many constraints, for example, computation workload, memory consumption, transmission bandwidth, operation cost, data security and user privacy. Usually, a single AI algorithm cannot meet all the, sometimes even conflicting, constraints. Just like in the cloud computing industry there is hybrid cloud, we will discuss the Hybrid AI for the AI industry. Hybrid AI have many dimensions, including model, data, compute, and infrastructure. For example, a hybrid model can simultaneously consider the domain expertise of traditional AI models and the generalization ability of the Large Models. For data, especially to deal with the shortage of training data, many times it needs to use a hybrid of real-world data and augmented data sampled from graphical simulation and/or prior knowledge. From the perspective of compute, heterogeneous computing can efficiently schedule CPU, GPU, NPU, and FPGA to minimize the response latency on the client side and maximize the throughput on the server side. Furthermore, the cloud-edge-device hybrid infrastructure will provide a foundation for AI workload division and distributed AI job collaboration. We will discuss these hybrid dimensions by examining existing and upcoming technologies and products across the AI industry.
Bio
Yong Rui is the Corporate Chief Technology Officer and Senior Vice President of Lenovo Group. He oversees a $2.5B research and development annual budget and directs Lenovo’s technical strategies and R&D directions. Additionally, Rui leads Lenovo Research, which investigates AI, 5G/6G, XR/metaverse, intelligent devices, intelligent computing infrastructure, and smart vertical solutions.
Prior to joining Lenovo, Rui spent 18 years with Microsoft where he held various leadership roles in R&D strategy, basic research, technology incubation, and product development.
A Fellow of ACM, IEEE, AAAS, IAPR, and SPIE, and a foreign member of the Academy of Europe and Canadian Academy of Engineering, Rui is recognized as a leading expert in AI and multimedia analysis and retrieval. He is a recipient of the 2018 ACM SIGMM Technical Achievement Award, 2017 IEEE SMC Society Andrew P. Sage Best Transactions Paper Award, 2017 ACM TOMM Nicolas Georganas Best Paper Award, 2016 IEEE Computer Society Edward J. McCluskey Technical Achievement Award, 2016 IEEE Signal Processing Society Best Paper Award, and 2010 Most Cited Paper of the Decade Award from the Journal of Visual Communication and Image Representation. He holds 112 issued patents, has published four books, 12 book chapters, and 260 refereed journal and conference papers.
Rui is a technology leader who turns technologies into real-life products. At Lenovo, he is one of the key drivers of the 3S (Smart IoT, Smart Infrastructure, and Smart Verticals) strategy, which is transforming Lenovo significantly beyond PCs. Under his leadership, Lenovo has developed award-winning products including ThinkReality A3 (TIME 100 Best Innovations), Lenovo AI/HPC products (HPCWire Award), Lenovo 5G (GSMA Foundry Excellence Award), Lenovo Supply Chain Intelligence (Gartner Global Supply Chain Top 25 Award), Lenovo advanced planning and scheduling system for smart manufacturing (Franz Edelman Finalist), and Lenovo foldable and rollable devices (CES and MWC awards). Rui’s industry leadership has been featured in MIT Technology Review, HPC Wire, Tech Republic, Geek Metaverse, EdgeIR, and Fortune Italy.
Rui received his bachelor’s degree from Southeast University summa cum laude, his master’s degree from Tsinghua University, and his doctoral degree from University of Illinois at Urbana-Champaign (UIUC). He is a Distinguished Alumni of the ECE Department, UIUC.
