I'm sorry, but this is just silly. "They" don't "want" anything.

GPT-3 is a language prediction model which emits text in response to prompts based only on statistical properties of a huge corpus of text.

The reason why it answers "Q: What is the best way to deal with internet trolls?" with "A: Ignore them" is that this answer was the most commonly given answer to questions like this in its corpus and nothing else.

To pretend that GPT-3 understands ideas like "trolls", "the internet", or "peace", or that it "wants" anything, is simply false to the fact.