Theme
Hidden Human Costs
Behind every model are data labelers, moderators and communities who were left out. The people who pay are rarely the ones in the room.
Episodes on this theme
The Invisible Labor Behind AI: A Content Moderator’s Fight for JusticeKauna MalgwiFormer Facebook content moderator (via Sama, Nairobi), leader of the Nigeria chapter of the Content Moderators Union, TIME100 AI 2024 honoree
AI for Humanity: A Conversation with Allison CohenAllison CohenSenior applied AI project manager, AI for Humanity team, Mila
Democratizing AI Education: A Conversation with Timothy from the AI Launch LabTimothyCo-Founder and Managing Director of the AI Launch Lab; member of the AI Advisory Board at Dawson College
AI and Cybercrime: The New Frontier in Digital ForensicsAlain FilottoDigital Evidence Specialist, Alphafox Forensics; retired RCMP member and former Chief Operations Officer, TCS Forensics
The Creator Economy, EdTech, and AI's Impact with Thomas FortoulThomas FortoulHost of the Thinking Tech podcast; helping an EdTech startup at Next.ai, MontréalIn their words
"AI never built itself." A former Facebook moderator on who really trained it.
The whole charade around AI, AI, we built AI, AI never built itself. So, and we are never acknowledged. If we keep quiet, it means AI really built itself, which is an absolute lie. People lost their life while building AI. We have colleagues that have committed suicide.Kauna Malgwi
Kauna explains why she keeps speaking out despite the personal cost of advocacy.
Same job, same platform: under $2 an hour in Nairobi, over $20 in the US.
The pay is significantly nothing because we're paid less than $2 per hour. And the same kind of job our colleagues are doing in the US, Europe, and on the other side of the world are being paid over $20 per hour. So you can see the discrepancy of payment.Kauna Malgwi
Kauna lists the biggest obstacles facing African content moderators.
If the job isn't harmful, why was therapy mandatory once a week?
So why are you providing counselors if you claim you are not aware that there's a mental health issues? If I should be in a supermarket selling groceries, would you provide a mental health counselor? Because you know it's not related.Kauna Malgwi
Kauna argues that Sama and Meta knew about the psychological harm because counseling was mandatory for moderators.
50 seconds per video. Even for a beheading. That was the quota.
No, you were forced to submit it in 50 seconds. So what were you asked to do? We were asked to watch the first 15 seconds, skip to the middle, watch 15 seconds, go to the end of the video, watch 15 seconds, and check the thumbnail and submit. So you can imagine if it's a beheading from an ISIS video. What does that do to you as a human? Because we are not robots.Kauna Malgwi
Kauna describes the productivity quotas Facebook moderators in Kenya worked under.
The people who label AI's data have been left out of the AI story for years.
I'm thinking about all the examples of the exploitation of data workers and the fact that that's sort of been swept under the rug for a really long time. But I really hope that people start to get more engaged in the field and the barriers to entry start lowering because it's going to continue to have such a dramatic effect on our world.Allison Cohen
In her closing message, Allison points to data workers' exploitation as something the AI narrative has hidden.
Measure clicks and you will build clickbait. The metric is a design decision.
And how you measure success will end up, of course, having a big impact on how you build these tools. If you're measuring click rate, you may optimize for, um, you know, click bait titles or manipulative tactics. And we don't want to do that.Allison Cohen
Allison explains why academic and Silicon Valley metrics like citations and click-through rates can distort how AI tools are designed.