Paper Title: Developing Chips for Tomorrow: You, Me, and ChatGPT

Paper Abstract: The big news of the year in computing is ChatGPT, a large language model that lets users easily apply AI to many everyday tasks. Areas from accounting through zoology all plan to use AI to create documents, drawings, marketing materials, computer programs, schematics, and more. However, ChatGPT and similar AI algorithms are very large and require huge computing resources. How can we create chiplets that will reduce both cost and execution time – and avoid enormous power bills and hotspots? Useful approaches include AI cores, efficient communication fabrics, more on-chip memory, more in-package components, more efficient die-to-die interfaces, and vertical chip stacking. Designers must improve and co-optimize multiple stages of chip development including die-partitioning, design workflows, system integration, IC packaging, testing, and manufacturing.

Paper Author: Jawad Nasrullah, CEO, Palo Alto Electron

Author Bio: Jawad Nasrullah is CEO of Palo Alto Electron, a startup focused on doing research on heterogeneous integrated circuits and developing 3D-ICs for performance computing. He was previously President, CTO, and Co-Founder of ZGlue, the creator of a platform for developing chiplets as well as a marketplace for distributing them. Before co-founding ZGlue, he was an engineer at Samsung Electronics, Intel, and Sun Microsystems. He earned a PhD in EE at Stanford, has 6 publications, and holds 14 patents.