Papers of the day   All papers

The De-democratization of AI: Deep Learning and the Compute Divide in Artificial

Comments

Abeba Birhane: There is a significant inequality in AI research according to a new paper https://arxiv.org/abs/2010.15581 Core findings: 1.Big Tech & elite universities have increased presence in top AI conf's since the rise of deep learning 2.Their increased presence has crowded out non-elite unis https://t.co/Zp4xtwdObF

12 replies, 478 likes


Erik Brynjolfsson: The De-democratization of AI: Deep Learning and the Compute Divide in AI Research by Nur Ahmed & @ImMuntasir "large tech firms and elite universities have increased participation in major AI conferences since...2012." HT: @tylercowen #AI @DigEconLab https://arxiv.org/abs/2010.15581

4 replies, 41 likes


Joshua Loftus: This paper is about a really important problem. One cause it mentions that I would emphasize more: the role of proprietary data. Most of the value of "AI" comes from the users and content moderators on massive platforms that generate/curate data

2 replies, 15 likes


JP Vergne: Really cool research by @iveybusiness school PhD candidate @NurAhmedB on the centralization of #AI firepower in the hands of a happy (and wealthy) few ⬇️

1 replies, 5 likes


Dagmar Monett: "[T]he divergence between ... large firms and non-elite universities is driven by access to computing power ... which we term as the "compute divide". [It] increases concerns around bias and fairness within AI technology, and presents an obstacle towards 'democratizing" AI." #AI

1 replies, 5 likes


Dr Catherine Breslin: There are increasing barriers to doing R&D in machine learning and AI: 1.Attracting talent & being able to pay the high wages that ML scientists command 2.Large amounts of good quality data for training models 3.Access to compute resources to effectively build large models

0 replies, 4 likes


Danil Mikhailov: Great thread from @Abebab and a very interesting paper. No doubt that compute power has a big effect on the growing disparity. My view is that access to compute power for academic and philanthropic work should be considered public infrastructure and subsidised by the state

0 replies, 3 likes


Mike Linksvayer: In the post-Moore era (since 2012), compute used for headline AI results doubling every 3.4 months rather than every 2 years, noticed in The De-democratization of AI: Deep Learning and the Compute Divide https://arxiv.org/abs/2010.15581

1 replies, 3 likes


alex zook: empirical analysis of the “compute divide” in AI. not encouraging for that whole “AI democratizes” narrative https://arxiv.org/abs/2010.15581 (via https://marginalrevolution.com/marginalrevolution/2020/11/is-ai-centralizing-research-influence.html)

0 replies, 3 likes


La Forge AI: [2010.15581] The De-democratization of AI: Deep Learning and the Compute Divide in Artificial Intelligence Research https://arxiv.org/abs/2010.15581

0 replies, 3 likes


Tactical Tech: The De-democratization of AI: Deep Learning and the Compute Divide in Artificial Intelligence Research https://arxiv.org/abs/2010.15581

0 replies, 2 likes


Jeff Lockhart: And since I teach and work in this area, it's also worth noting that very few universities have the $ to build their own HPC infrastructure to teach students on. Free infrastructure from cloud providers is the only alternative to this: https://twitter.com/Abebab/status/1322229024387727360?s=19

1 replies, 2 likes


Roberto Maestre: The De-democratization of AI: Deep Learning and the Compute Divide in Artificial Intelligence Research https://arxiv.org/abs/2010.15581 https://t.co/UmvRECDRtF

0 replies, 1 likes


Jenny Andrew 🏴‍☠️: Depressingly predictable, and self-evidently harmful to the science...

0 replies, 1 likes


Philosophy of Futures Studies: https://arxiv.org/abs/2010.15581 "The De-democratization of AI: Deep Learning and the Compute Divide in Artificial Intelligence Research" #philofutures #ArtificialIntelligence #fridaymorning https://t.co/tAj48j2y0u

0 replies, 1 likes


James Cuff: The digital divide continues. The haves and the have nots isn’t anything new. The democratization of access is a big dealio in #hpc and #ai. Detailed analysis in this paper. Thanks for sharing Abeba, a must read for any center director / facility mangr, especially national ctrs!

0 replies, 1 likes


srisatish: The De-democratization of AI: Deep Learning and the Compute Divide in Artificial Intelligence Research https://arxiv.org/abs/2010.15581

0 replies, 1 likes


fully autonomous landback communities: @Astro_Erik The reason that it keeps coming up, is because AI research and computational sciences fields are themselves broken - the people who do the research, at which schools, and with what financial backing. See here: https://arxiv.org/abs/2010.15581 and here: https://arxiv.org/abs/2009.14258

1 replies, 1 likes


Julius Frost: Paper: The De-democratization of AI: Deep Learning and the Compute Divide in Artificial Intelligence Research https://arxiv.org/abs/2010.15581

1 replies, 0 likes


Content

Found on Oct 30 2020 at https://arxiv.org/pdf/2010.15581.pdf

PDF content of a computer science paper: The De-democratization of AI: Deep Learning and the Compute Divide in Artificial