2019/11/12 by Jeffrey Dean, Dean, Jeffrey · 4 voices
Engineering · #Advancements in Semiconductor Devices and Circuit Design #Ferroelectric and Negative Capacitance Devices #Semiconductor materials and devices
paper · pdf · doi:10.48550/arxiv.1911.05289
The past decade has seen a remarkable series of advances in machine learning,\nand in particular deep learning approaches based on artificial neural networks,\nto improve our abilities to build more accurate systems across a broad range of\nareas, including computer vision, speech recognition, language translation, and\nnatural language understanding tasks. This paper is a companion paper to a\nkeynote talk at the 2020 International Solid-State Circuits Conference (ISSCC)\ndiscussing some of the advances in machine learning, and their implications on\nthe kinds of computational devices we need to build, especially in the\npost-Moore's Law-era. It also discusses some of the ways that machine learning\nmay also be able to help with some aspects of the circuit design process.\nFinally, it provides a sketch of at least one interesting direction towards\nmuch larger-scale multi-task models that are sparsely activated and employ much\nmore dynamic, example- and task-based routing than the machine learning models\nof today.\n