4 ms·
But it doesn't need to be fully automated. Automate 999 out of 1000 (probably quite easily) and flag the rest for human review. Surely we've got enough GPUs now
by chillydawg 8y ago
But it doesn't need to be fully automated. Automate 999 out of 1000 (probably quite easily) and flag the rest for human review. Surely we've got enough GPUs now to train a neural network smart enough to do that.
- intended 8y agoAutomation of moderation IS very dangerous. If there is one thing people need to realize, is that automated moderation is inherently dangerous. People think of news and informaiton online. but its all just Content. CONTENT can be anything, fake news, lies, or fighting against an unjust leader, rebellion, fighting for the environment and more. Tools that effectively get better at moderating content do so for ALL content. Any tool that you use to cut bad will is always useable to harm good. -------- So far I have good reason to assume that intent is not paresable from content online, so human beings should be relatively safe online. Figuring out INTENT, is the holy grail of moderation. If you know why someone said something, you know whether its meant as Nazi apologia or whether its no malicious historical context. As long as we don't figure out a way to automate that, private thoughts are safe.