Skip to content

Latest commit

 

History

History
30 lines (20 loc) · 2.94 KB

index.md

File metadata and controls

30 lines (20 loc) · 2.94 KB

SECure: A Social and Environmental Certificate for AI Systems

A project by Abhishek Gupta, Camylle Lanteigne, and Sara Kingsley

SECure has so far been presented at the Orchestrating Knowledges Symposium (CANSEE), the Extraction: Tracing the Veins (Massey University), the Global Challenges in Economics and Computation Conference 2020, MD4SG 2020 Workshop, and the ICML Deploying and Monitoring Machine Learning Systems workshop 2020.

A Note on this Project

This project is currently in its infancy, and we (the authors) plan to continue to expand on this both theoretically and concretely as a framework to promote environmentally and socially responsible ML. We invite you to submit feedback through this form and really appreciate your contribution into making this a more useful tool for the community!

Abstract

In a world increasingly dominated by AI applications, an understudied aspect is the carbon and social footprint of these power-hungry algorithms that require copious computation and a trove of data for training and prediction. While profitable in the short-term, these practices are unsustainable and socially extractive from both a data-use and energy-use perspective. This work proposes an ESG-inspired framework combining socio-technical measures to build eco-socially responsible AI systems. The framework has four pillars: compute-efficient machine learning, federated learning, data sovereignty, and a LEEDesque certificate. Compute-efficient machine learning is the use of compressed network architectures that show marginal decreases in accuracy. Federated learning augments the first pillar's impact through the use of techniques that distribute computational loads across idle capacity on devices. This is paired with the third pillar of data sovereignty to ensure the privacy of user data via techniques like use-based privacy and differential privacy. The final pillar ties all these factors together and certifies products and services in a standardized manner on their environmental and social impacts, allowing consumers to align their purchase with their values.

ArXiv Pre-print Paper

Find it here! https://arxiv.org/abs/2006.06217

SECure was featured on VentureBeat

Read it here! https://venturebeat.com/2020/06/12/researchers-propose-framework-to-measure-ais-social-and-environmental-impact/

SECure was featured in a video by Alex Castrounis

Watch it here! https://www.youtube.com/watch?v=zUKsRrf2KIQ

SECure was presented at the Extraction: Tracing the Veins Conference (Massey University)

Watch the presentation here! https://youtu.be/q2ALIe3VezE

Here is an illustration of SECure (made by @playthink on Twitter)

alt text

Contact

Reach out to the authors here: abhishek [at] montrealethics [dot] ai | skingsle [at] cs [dot] cmu [dot] edu