 |
THE IMPACT OF CHIPS ON AI & ETHICS By Berit Anderson
FiRe is back! Future in Review returns, live, at the beautiful Terranea Resort in Palos Verdes, CA, November 6-9. Join old friends and new at this life-changing event. Register now and get your Earlybird ticket for just $3,900 - a savings of $2,000. See more in "Upcoming Events." _____ How will advancements in chips drive the deployment of AI? And what impact will that have on society's ability to deploy AI ethically? These are the questions we asked chip experts Jon Peddie, Oskar Mencer, and Matt Keener last week at the first roundtable discussion in our latest Future in Review Action Tank series, "Chips & the Ethical Deployment of AI." We wanted to better understand the intersection between hardware and machine learning, how new chip architectures will shape the future of artificial intelligence, and what moral hazards and opportunities might lie ahead. So we called in a trio of the smartest folks we know on the topic to help us understand the playing field:
Jon Peddie, PhD, is a recognized pioneer in the graphics industry, is president of Jon Peddie Research, and has been named one of the world's most influential analysts. He is an ACM Distinguished Speaker, an IEEE Distinguished Visitor, and an IEEE Computer Society Distinguished Contributor and charter member. He lectures at numerous conferences and universities on topics about graphics technology and the emerging trends in digital media technology. Formerly president of Siggraph Pioneers, Jon serves on the advisory boards of several conferences, organizations, and companies and contributes articles to numerous publications. In 2015, he was given the Lifetime Achievement Award from the CAD Society. Jon has published hundreds of papers, written 10 and co-authored a half-dozen technical books (including Augmented Reality: Where We All Will Live, Ray Tracing: A Tool for All, and his latest, the three-book series The History of the GPU), as well as two fictional titles.
Oskar Mencer received a PhD from Stanford University following studies at the Computer Systems Laboratory with Prof. Michael J Flynn. He had received an earlier degree in computer engineering from the Technion in Israel. Oskar worked for Rockwell, DIGITAL, Hitachi Central Research Laboratories in Japan, and Bell Labs before starting Maxeler Technologies in 2003 and enjoying the variety of holding the CEO position until today, even through Groq's acquisition of Maxeler in 2022. Besides Maxeler, Oskar is chair of Haptomass, an Imperial College London spinout on 2D mass spectroscopy; a member of the Board of Governors at the Technion; and a member of Academia Europaea.
Matt Keener is the lead hardware designer at Pattern Computer Inc. (PCI), where he is responsible for hardware design and architecture. Matt is an expert in embedded systems and logic design, with a focus on hardware architecture design, bring-up, and debugging. He has extensive experience in designing logic in application-specific integrated circuits (ASICs) and field-programmable gate arrays (FPGAs), designing printed circuit boards (PCBs) for high-speed digital signaling, and firmware for control of both. Matt's experience spans application-specific timing and control logic up to system-on-chip designs that include embedded processors. He also has research experience in semiconductor physics and biomedical engineering, with a focus on biosensors. Prior to PCI, Matt worked at IBM, National Instruments, Ohio State University, and GE Aircraft Engines. Throughout his career, he has developed a reputation for providing well-researched solutions. He has led and participated in various cross-functional projects, working across continents with team members with various levels of experience. As electronics sourcing company Vyrian wrote in a recent report: Artificial Intelligence (AI) is transforming the semiconductor industry, affecting everything from chip design and development to production. AI-driven chips are revolutionizing how engineers design and create them while also enabling unprecedented capabilities such as big data analytics. With chip architectures rapidly evolving, companies in the semiconductor industry need to understand just how this technology is impacting the process of designing, developing, and producing modern chips so they can stay ahead of their competition. Personally, I had just one problem: as a non-expert on chip architecture, I needed help understanding the basics of machine-learning chips. You don't ask the president of the United States to explain the functions of the three branches of government to you. And neither did I want to waste these folks' time reviewing basic chip structures. So, in the spirit of our roundtable series, I decided to lean on artificial intelligence. It was the perfect application for ChatGPT. This wasn't a high-stakes environment - after all, I would be relying on human experts to tell me what was really going on. And I was generally aware of the limitations of ChatGPT - for example, it's been programmed by its developers to avoid explicit or offensive content. (Yawn.) It's loath to make value judgments, which the ethical application of AI demands. And its training dataset is composed of primarily public information, based on news and academic research, and cuts off in 2021. What's' more, it sometimes hallucinates. So, I wasn't likely to learn anything truly new or revolutionary from my artificial advisor. But as an efficient source of baseline public information, it would do the trick quite nicely.
A ChatGPT-Powered Primer on Machine-Learning Chips |
