|
|
|
|
|
by albertsun
2225 days ago
|
|
It's only a tradeoff if you assume that the process must be automated and operate with a minimum of actual human scrutiny or consideration. Any qualified human review would immediately be able to distinguish and vastly improve both the false negative and false positive rates. But because these tech companies devalue that kind of work they aren't willing to invest in it. |
|
1. There's a real social cost associated with constantly subjecting human reviewers to traumatizing content. It can lead to devastating mental health problems. (some relevant articles: https://www.theverge.com/2019/2/25/18229714/cognizant-facebo..., https://www.telegraph.co.uk/technology/2019/12/16/youtube-mo...)
2. Humans make mistakes as well, especially as volume of content increases and policies get more nuanced.
3. There's way more violating content then you could ever imagine. As just one example, YouTube reported having to remove >1.8 BILLION comments for spam or other policy violations in 2019 (https://transparencyreport.google.com/youtube-policy).
4. The large tech companies spend billions on human moderation as it is. YouTube alone for example has >10,000 full time human reviewers. (https://youtube.googleblog.com/2017/12/expanding-our-work-ag...)