The 'dark forest' of the web expands with generative AI

Maggie Appleton draws on Yancey Strickler's 'dark forest' metaphor to describe a web already overrun with bots and automated content. Now, with generative models like ChatGPT, Midjourney, and Stable Diffusion, the forest is poised to expand dramatically. These tools can produce coherent text and images at scale, promising to flood the web with generic, generated content—from SEO articles to social media posts—that could drown out authentic human expression.

Appleton argues that the key differentiator between humans and language models is access to shared reality. Humans can 'triangulate' with the world, grounding their words in embodied experiences and cultures. In contrast, language models are linguistic prediction systems that regurgitate training data. To survive, humans must cultivate originality, critical thinking, and niche knowledge, and lean into the informal, ever-evolving 'la parole' of everyday language.

However, these strategies may not be equally available to everyone. People with disabilities, those in remote areas, or with caretaking responsibilities may find it harder to verify their humanity in an offline-first world. The stakes hinge on whether we can foster authentic connections and intellectual life without resorting to exclusionary or impractical verification schemes.