Chenhao Tan

The slides of talks at the Human+AI conference is available here: https://chenhaot.com/talks.html

Berkeley convincingly shows that the objective function shouldn’t be an afterthought and people have diminishing sensitivity of error.

Berkeley Dietvorst argues for the important for AI objectives to align with human preferences.

Jenn provides recommendations for future guidelines/systems for dataset documentation: https://twitter.com/ChenhaoTan/status/1586005686932148224/photo/1

(image uploading seems challenging here.)

Jenn covers the details of disaggregated evaluation and argues that these details matter!

Jenn Wortman Vangaun kicks off the human+AI conference by talking about the disconnect between research on fairness and industry practitioners.