Green AI

Roy Schwartz, Jesse Dodge, Noah A. Smith, Oren Etzioni

Research output: Contribution to journalArticlepeer-review

519 Scopus citations

Abstract

Researchers suggest that creating efficiency in artificial intelligence (AI) research will decrease its carbon footprint and increase its inclusivity as deep learning study should not require the deepest pockets. The field of AI has reported remarkable progress on a broad range of capabilities including object recognition, game playing, speech recognition, and machine translation. Much of this progress has been achieved by increasingly large and computationally intensive deep learning models. An important study has estimated the carbon footprint of several NLP models and argued this trend is both environmentally unfriendly and prohibitively expensive, raising barriers to participation in NLP research, which is known as Red AI. An alternative is Green AI, which treats efficiency as a primary evaluation criterion along with accuracy. To measure efficiency, we suggest reporting the number of floating-point operations required to generate a result. Green AI research will decrease AI’s environmental footprint and increase its inclusivity.

Original languageAmerican English
Pages (from-to)54-63
Number of pages10
JournalCommunications of the ACM
Volume63
Issue number12
DOIs
StatePublished - 17 Nov 2020

Fingerprint

Dive into the research topics of 'Green AI'. Together they form a unique fingerprint.

Cite this