Papers of the day   All papers

The Computational Limits of Deep Learning

Comments

Denny Britz: The authors of this paper analyzed 1,058 arXiv papers and plotted various benchmarks against the increase in compute requirements, arguing that the current progress is largely driven by more compute and may become unsustainable soon: https://arxiv.org/abs/2007.05558 https://t.co/hWCVhK11GY

10 replies, 283 likes


MIT CSAIL: MIT study: with deep learning's carbon footprint growing exponentially. neural nets will survive only if they, and the hardware they run on, become radically more efficient. Paper: https://arxiv.org/pdf/2007.05558.pdf More: https://news.mit.edu/2020/shrinking-deep-learning-carbon-footprint-0807 Work w/Yonsei U & @unb_oficial #ECCV2020 https://t.co/zFTFdU1Vh6

1 replies, 142 likes


Anna Rogers: In #GreenAI news: https://arxiv.org/abs/2007.05558 TLDR: analysis of 1,058 papers showed that progress in CV, QA, NER, and MT was "strongly reliant" on increases in computing power, and we literally can't afford that kind of progress anymore. Work by @ProfNeilT @KGreenewald et al. /1 https://t.co/IatGSUJW0i

2 replies, 56 likes


Trishank Karthik Kuppusamy: Very interesting... @spyrosmakrid https://arxiv.org/abs/2007.05558 https://t.co/cyAlZUOc0e

2 replies, 28 likes


Sam Gershman: This is a missed opportunity to herald the arrival of the "AI summer" when C02 emissions from deep learning push global temperatures past the brink. https://arxiv.org/abs/2007.05558

0 replies, 28 likes


Hacker News: The Computational Limits of Deep Learning https://arxiv.org/abs/2007.05558

0 replies, 23 likes


Gavin Baker: 2) Link to the underlying academic paper and screenshot: https://arxiv.org/abs/2007.05558 https://t.co/IGh6pyQhro

5 replies, 18 likes


Robin Hanson: "Deep Learning [progress] in 5 prominent… areas… strongly reliant on increases in computing power. Extrapolating forward this reliance reveals… progress along current lines is rapidly becoming economically, technically, & environmentally unsustainable." https://arxiv.org/abs/2007.05558

5 replies, 12 likes


Arthur Charpentier: "The Computational Limits of Deep Learning" https://arxiv.org/abs/2007.05558

0 replies, 11 likes


Natesh Ganesh: Looking forward to digging into this paper - https://arxiv.org/abs/2007.05558 https://t.co/qbMiGUvvol

0 replies, 11 likes


Neil Thompson: A.I.'s most exciting successes have come from Deep Learning, which has a voracious appetite for computing power. That's quickly becoming unsustainable, as I show in my new paper with @KGreenewald (@MITIBMLab), Keeheon Lee (@yonsei_u) and Gabriel Manso: https://arxiv.org/abs/2007.05558

0 replies, 9 likes


Matt Challacombe: The Computational Limits of Deep Learning https://arxiv.org/abs/2007.05558 "Another option is to abandon deep learning entirely and concentrate on other forms of machine learning that are less power hungry."

0 replies, 4 likes


Estesis: The Computational Limits of Deep Learning Thompson et al.: https://arxiv.org/abs/2007.05558 #ArtificialIntelligence #DeepLearning #MachineLearning https://t.co/yLx7pzvuuc

0 replies, 4 likes


urmas pitsi: Some believe that at 10^19 GFlops we'll hit singularity and/or run ancestor simulations. Others think that at best we'll achieve 5% error rate on ImageNet. Am I missing something? :) @ykilcher,@lexfridman, @KGreenewald, @ProfNeilT

1 replies, 4 likes


Lum: “The Computational Limits of Deep Learning” - https://arxiv.org/pdf/2007.05558.pdf

1 replies, 4 likes


akira: https://arxiv.org/abs/2007.05558 The paper suggests that Deep Learning has improved the performance of many tasks by using vast amounts of computational power, but this suggests that it may stall depending on hardware, as the computational power required is growing larger and larger. https://t.co/0MbC7v8yH2

0 replies, 4 likes


Jacques Thibodeau: Something to think about.

0 replies, 4 likes


Michael P. Frank: If they used reversible computing they wouldn’t be facing limits!

0 replies, 4 likes


urmas pitsi: The Computational Limits of Deep Learning https://arxiv.org/abs/2007.05558

0 replies, 3 likes


Ehud Reiter: Aberdeen NLP reading group for 14 Oct The Computational Limits of Deep Learning https://arxiv.org/abs/2007.05558

0 replies, 3 likes


Brundage Bot: The Computational Limits of Deep Learning. Neil C. Thompson, Kristjan Greenewald, Keeheon Lee, and Gabriel F. Manso http://arxiv.org/abs/2007.05558

1 replies, 3 likes


Hacker News 50: The Computational Limits of Deep Learning https://arxiv.org/abs/2007.05558 (https://bit.ly/3g3omnp)

0 replies, 2 likes


(((JReuben1))): The Computational Limits of Deep Learning https://arxiv.org/abs/2007.05558

0 replies, 2 likes


Curt Langlotz: "Deep learning needs computational power for image classification and other problems. If progress continues at current pace, computational requirements will rapidly become technically and economically prohibitive" The Computational Limits of Deep Learning http://arxiv.org/abs/2007.05558 https://t.co/IHvtxbt7YF

0 replies, 1 likes


Jason H. Moore, PhD: The Computational Limits of Deep Learning https://arxiv.org/abs/2007.05558 HT @curtlanglotz #deeplearning

0 replies, 1 likes


DLビギナ: https://twitter.com/learn_learning3/status/1289128247255379975?20311

0 replies, 1 likes


Hacker News 20: The Computational Limits of Deep Learning https://arxiv.org/abs/2007.05558 (https://bit.ly/3g3omnp)

0 replies, 1 likes


mauro_iker: It's time for #GreenAI Read this paper for a practical solution by making efficiency an evaluation criterion for research alongside accuracy and related measures 👇 https://arxiv.org/abs/1907.10597 #ArtificialInteligence #DeepLearning #ClimateChange #MachineLearning

0 replies, 1 likes


Phil Frana: The Computational Limits of Deep Learning https://arxiv.org/pdf/2007.05558.pdf

0 replies, 1 likes


Content

Found on Jul 18 2020 at https://arxiv.org/pdf/2007.05558.pdf

PDF content of a computer science paper: The Computational Limits of Deep Learning