Engineering and Technology

An analog-AI chip for energy-efficient speech recognition and transcription

S. Ambrogio, P. Narayanan, et al.

Discover how a groundbreaking analog-AI chip with 35 million phase-change memory devices achieves remarkable energy efficiency, boasting performance levels up to 12.4 TOPS/W. This technology not only ensures software-equivalent accuracy for keyword spotting but also approaches it for more extensive models, demonstrating significant potential for the future of speech recognition and transcription—all developed by S. Ambrogio, P. Narayanan, A. Okazaki, and their esteemed colleagues at IBM Research.

00:00

Playback language: English

Index

Abstract

Large AI models, while achieving high accuracy, suffer from poor energy efficiency on conventional processors. Analog in-memory computing (analog-AI) offers improved energy efficiency by performing parallel matrix-vector multiplications. This paper presents an analog-AI chip with 35 million phase-change memory devices, achieving up to 12.4 TOPS/W performance. The chip demonstrates full software-equivalent accuracy for a small keyword-spotting network and near-software accuracy on the larger MLPerf RNNT, showcasing the potential of analog-AI for energy-efficient speech recognition and transcription.

Publisher

Nature

Published On

Aug 23, 2023

Authors

S. Ambrogio, P. Narayanan, A. Okazaki, A. Fasoli, C. Mackin, K. Hosokawa, A. Nomura, T. Yasuda, A. Chen, A. Friz, M. Ishii, J. Luquin, Y. Kohda, N. Saulnier, K. Brew, S. Choi, I. Ok, T. Philip, V. Chan, C. Silvestre, I. Ahsan, V. Narayanan, H. Tsai, G. W. Burr

Related Publications

Explore these studies to deepen your understanding of the subject.

Chemistry

DECIMER.ai: an open platform for automated optical chemical structure identification, segmentation and recognition in scientific publications

K. Rajan, H. O. Brinkhaus, et al.

Medicine and Health

An automatic speech analytics program for digital assessment of stress burden and psychosocial health

A. M. Y. Chu, B. S. Y. Lam, et al.

Biology

Design of a recombinant asparaginyl ligase for site-specific modification using efficient recognition and nucleophile motifs

J. Tang, M. Hao, et al.

Engineering and Technology

Fast encirclement of an exceptional point for highly efficient and compact chiral mode converters

X. Shu, A. Li, et al.

Listen, Learn & Level Up

Over 10,000 hours of research content in 25+ fields, available in 12+ languages.

No more digging through PDFs, just hit play and absorb the world's latest research in your language, on your time.

listen to research audio papers with researchbunny