The Expanding Dark Forest and Generative AI
Jan 4, 2023

Maggie Appleton posted this article, which made it to the front page of Hacker News today:
https://maggieappleton.com/ai-dark-forest

Seems that this idea is certainly resonating with the very online subset of techies that read HN, which is somewhat surprising to me.

I love the metaphor and the illustration, I chuckled a bit while reading it. That being said, sometimes models and metaphors like this are posed because they make us feel more comfortable about the decisions we make, not because they accurately model what's going to actually happen. The HN crowd loves the cozy web because that's what HN is - a highly curated version of Reddit.

So what will actually happen? Based upon how much easier it is to detect ai-generated content than people believe, the dark forest is going to put money into detecting and flagging automated content - for example with StackOverflow now prohibiting automated content.

Also people are going to figure out soon enough that Large Language Models create huge errors, so people are going to have to correct the errors in the output, at a minimum, before posting. Marketers and bloggers are going to figure out that they have to significantly modify LLM-generated text or else they get flagged or shadow-banned. So yeah...there's already probably a ton of auto-content out there, but it's easier to detect than people think.

© Manifold Markets, Inc.Terms + Mana-only TermsPrivacyRules