What It’s Like to Be a Data Labeler Training AI

Описание: I recently traveled to Kenya for a journalism and AI conference. While I was there, I really wanted to meet with Michael Geoffrey Asia, the secretary general of the Data Labelers Association. Data Labeling is a huge job in Kenya. Data labelers are the people who train AI, and who also work on ensuring the outputs are accurate. In some cases, data labelers are themselves pretending to be AI, in order to train AI. Often, data labelers don’t know exactly what they’re working on, because the work usually goes through a platform, a subcontractor, or a combination of both. So basically they can be presented with a backend where they’re asked to perform tasks or answer questions; in some cases their answers may be presented in real time as AI.

Data labeling is notoriously brutal and underpaid work. Workers sometimes earn as little as a few dollars a day, work under algorithmic management, and, because they’re sometimes trying to train AI what not to do or show, they are often shown graphic, violent, or sexual content for hours at a time. It’s kind of similar to content moderation jobs, and lots of people do both data labeling and content moderation, or switch back and forth between the industries. It’s such a big thing in Kenya that I mentioned it to the driver who took me to meet Michael for this interview, and she told me that she too was a data labeler, as are many of her friends.

Michael has since become critical at the Data Labelers Association, a group that is fighting to organize people who do data labeling work and who is advocating for better working conditions, higher pay, and more protections for data labelers. I met Michael at a coworking space in Nairobi in a very tiny room, so I’m not on camera after this, but here’s my conversation with Michael.

The Emotional Labor Behind AI Intimacy by Michael Geoffrey Asia: https://data-workers.org/wp-content/u...

0:00 - Intro
5:39 - What Is Data Labeling?
7:11 - The Growth of the Data Labeling Industry in Kenya
8:29 - Michael’s Introduction to Data Labeling
10:15 - Pressures of Data Labeling
11:23 - Inside the Work of Sama
12:12 - Annotating Graphic Images
15:25 - Training AI Companions
17:38 - Cultural Differences Between Kenyans and Chatbot Users
19:10 - Explicit Chabot Conversations
19:25 - The Mental and Physical Toll of Data Labeling
21:41 - Mental Health Support for Data Labelers
23:09 - The Exploitative Nature of Data Labeling
30:03 - The Data Labelers Association
31:19 - Reforming Data Labeling
33:09 - Can Data Labeling Be Reformed?
35:15 - Is There Such Thing as Ethical Data Labeling?
38:02 - Sama’s Client Companies
40:53 - How Common Is Data Labeling in Kenya?
41:50 - Witnessing Violent Imagery As A Data Labeler
42:51 - NDAs
43:54 - Training AI to Replace You
47:20 - Data Labelers Association Outreach

This is a production of 404 Media, a journalist-owned tech website. Learn more and subscribe at: htttps://404media.co

Listen to our weekly podcasts:

Apple Podcasts: https://podcasts.apple.com/us/podcast...
Spotify: https://open.spotify.com/show/0F3oY47...
Google Podcasts: https://podcasts.google.com/feed/aHR0...

Become a paid subscriber for access to bonus content: https://404media.co/membership