2018



A common practice for machine learning engineering is to use Amazon Mechanical Turk to label datasets.  AMT is a marketplace for work that requires human intelligence like audio transcription, translation, identifying sentiments, categorising images etc. But, who are those people who annotate images and participate in large psychological predictive experiments?

100k happy moments were recently crowdsourced on Amazon Mechanical Turk where workers were asked to describe happy moments of the past 24 hours.

I made a Twitterbot that tweets those moments. Happy Amazon Turker visualises how the boundaries paid labor, data collection for AI systems and personal information are blurred.

Data: https://rit-public.github.io/HappyDB/ (a corpus of 100,000 crowd-sourced happy moments)

Twitter bot: https://twitter.com/amznturkerhappy