[ExI] Training GPT-3 to play nice - it's difficult!

[Ben Zaiboc]

On 02/11/2020 17:09, BillK wrote:
> How to make a chatbot that isn’t racist or sexist
> Tools like GPT-3 are stunningly good, but they feed on the cesspits of
> the internet. How can we make them safe for the public to actually
> use?

Wrong question, I reckon.

This is basically asking "how can we enforce a specific ideology on 
software?". Making software PC is basically censorship.

It's a bit like wanting to make sure artists don't portray naked flesh, 
or writers don't use swear words.

What the right question is, I'm not exactly sure. I'm inclined to say, 
though, that the problem doesn't lie with the software.

-- 
Ben Zaiboc