

Abstract: With the rapid rise of AI in various fields, we naturally wonder whether it can help or even replace analog designers. We first ponder a general question: how can AI become a competent analog designer? We postulate that AI must climb a five-level pyramid. At the foundation, AI must analyze circuits correctly. This ability appears to be maturing in some AI tools, especially if it simply requires writing and solving equations. The second level relates to intuition: can AI understand a structure well enough to predict its behavior as we vary certain parameters? Possessed by good analog engineers, this key skill guides the design process, distinguishing them from “SPICE monkeys.”
At the third level, AI must be able to design a given circuit topology according to certain specifications. The task of training AI for this purpose presents formidable challenges. The fourth level extends this ability to multiple topologies for the same function, seeking a global optimum. The fifth level of the pyramid calls for innovation in circuits, in a manner similar to what humans have done for a century.
In this presentation, we evaluate the present capabilities of AI and probe its conceptual understanding beyond what it can find online. We begin with analysis questions and then extend our effort toward circuit design.
Registration link: https://ieee.webex.com/weblink/register/mcdc33da1b1d485f5191e37fdec56531f
Bio: Behzad Razavi is Professor of Electrical Engineering at UCLA, where he conducts research on analog and RF integrated circuits. Prof. Razavi has served as an IEEE Distinguished Lecturer and has published more than 200 papers and nine books. He has received nine IEEE PBest paper Awards and six teaching and education awards, and his books have been published in eight languages. He received the IEEE Pederson Award in Solid-State Circuits and was recognized as a top author in the 50-year and 75-year histories of the IEEE International Solid-State Circuits Conference. He is a member of the US National Academy of Engineering.