Hacker News Reader

The outputs of the system will reflect the choices that humans made about the inputs and processing. It makes a whole lot of sense to have humans taking a look at the outputs and making sure they are reasonable.

You obviously don't need humans reviewing the outputs of simple mechanisms, but as your mechanisms get more complex and opaque, the pretense that they are objective becomes ridiculous.

This. You only have to take a look at what the internet did to Microsoft's learning chatbot to see why humans are a good idea.

OTOH I only have to look at Microsoft's chatbot to see why humans are a bad idea, too ;-)

But if computers and humans are both incredibly stupid, at least they're stupid in different ways? And I do agree with his point that both can end up with biases.

I think you haven't heard of 4chan. What do you think happens when somebody makes an easily gameable system with no supervision and puts in on the Internet?