5 ms·
"To avoid this problem, according to Time, OpenAI engaged an outsourcing company that hired contractors in Kenya to label vile, offensive, and potentially illeg
by 1vuio0pswjnm7 4y ago
"To avoid this problem, according to Time, OpenAI engaged an outsourcing company that hired contractors in Kenya to label vile, offensive, and potentially illegal material that would then be included in the training data so that the company could create a tool to detect toxic information before it could reach the user. Time reported that some of the material "described situations in graphic detail like child sexual abuse, bestiality, murder, suicide, torture, self-harm, and incest." The contractors said that they were supposed to read and label between a hundred and fifty and two hundred and fifty passages of text in a nine-hour shift. They were paid no more than two dollars an hour and were offered group therapy to help them deal with the psychological harm that the job was inflicting. The outsourcing company disputed those numbers, but the work was so disturbing that it terminated its contract eight months early. In a statement to Time, a spokesperson for OpenAI said that it "did not issue any productivity targets," and that the outsourcing company "was responsible for managing the payment and mental health provisions for employees," adding that "we take the mental health of our employees and those of our contractors very seriously.""
- rvba 4y agoI wonder how much better the labeling of everything (not just the bad stuff) would be if it wasnt outsourced to the lowest bidder
- beaned 4y agoIs there a specific capacity in regards to labeling that is enabled by more money? I can see it for like.. heart surgery. But don't most of us know what things are to be called when we see them? ChatGPT seems to be pretty good at knowing what bad stuff is.
- rvba 4y agoI assume that it would lead to higher quality labels - > higher quality output. Analogy to the "Garbage in, garbage out" rule.
- czbond 4y agoThe goal is absolutely logical (training the model with data that is undesirable). It is the method you are questioning? I guess I would have just labeled them "terrible" or something in a bulk assignment to not have people observe them.
- 1vuio0pswjnm7 4y agoThis passage comes straight from the OP. On the positive side, 21 up 0 down. Readers appear to see the issue. On the negative side, several "tone deaf" commenters that appear to have a blind spot for human suffering. This is not like a clinical trial of a therapeutic. It is not even like animal testing of a consumer product. OpenAI knows the material is harmful. So-called "tech" companies like Facebook and Google are engaging in this sort of practice every day, paying people some embarassingly low wage, usually through contractors, to subject themselves to psychologically harmful content so the company can sell more online advertising services.