Empirical analysis from Roy Schwartz (Hebrew University of Jerusalem) and Jesse Dodge (AI2) suggests the AI research community has paid relatively little attention to computational efficiency. A focus on accuracy rather than efficiency increases the carbon footprint of AI research and increases research inequality. In this episode, Jesse and Roy advocate for increased research activity in Green AI (AI research that is more environmentally friendly and inclusive). They highlight success stories and help us understand the practicalities of making our workflows more efficient.

Leave us a comment

Changelog++ members save 5 minutes on this episode because they made the ads disappear. Join today!

Sponsors:

• The Brave Browser – Browse the web up to 8x faster than Chrome and Safari, block ads and trackers by default, and reward your favorite creators with the built-in Basic Attention Token. Download Brave for free and give tipping a try right here on changelog.com.
• Code-ish by Heroku – A podcast from the team at Heroku, exploring code, technology, tools, tips, and the life of the developer. Check out episode 98 and episode 99 for insights on the ethical and technical sides of deep fakes. Subscribe on Apple Podcasts and Spotify.
• Knowable – Learn from the world’s best minds, anytime, anywhere, and at your own pace through audio. Get unlimited access to every Knowable audio course right now. Click here to check it out and use code CHANGELOG for 20% off!

Featuring:

• Roy Schwartz – Twitter, lab-huji.github.io">Website
• Jesse Dodge – Twitter
• Chris Benson – Twitter, GitHub, LinkedIn, Website
• Daniel Whitenack – Twitter, GitHub, Website

Show Notes:

• Green AI article in the communications of the ACM
• Training a single AI model can emit as much carbon as five cars in their lifetimes
• Fine-Tuning Pretrained Language Models: Weight Initializations, Data Orders, and Early Stopping
• Parameter-Efficient Transfer Learning for NLP
• Reproducibility at EMNLP 2020

Something missing or broken? ai-124.md">PRs welcome!