If you've used ChatGPT, you may have noticed that it's relatively decent at avoiding toxicity. Is it because of some magical AI trick?
Not really.
The real heroes are the ghostworkers who have worked tirelessly to detect toxicity in text containing "child sexual abuse, bestiality...torture, self harm, and incest" (many of which are illegal by US laws).
How much were they paid for this mentally scarring task?
#ChatGPT #AI #ghostwork #datawork #labor #ResponsibleAI #ML #gpt3
1/3
Did you know #OpenAI paid outsourced Kenyan workers only $1.32-$2 per hour for this extremely burdensome task?
While the datawork done by these ghostworkers made ChatGPT less toxic, the work left people mentally scarred.
However, the moment the company got whiff of the Time's investigation, they terminated the contract, leaving these human beings high and dry.
https://time.com/6247678/openai-chatgpt-kenya-workers/
2/3
When we marvel at #AI advances, let us also reflect -- what type of labor made this possible? Was it exploitative?
Much like massive physical monuments were built off the backs of exploited people, the massive digital models are often not that different.
#ChatGPT #AI #ghostwork #datawork #labor #ResponsibleAI #ML #gpt3
3/3
Upol Ehsan
Tim Truxell