404 Future Not Found

We'll soon see such a flood of Generative-AI-Generated "content" — or what I call GAG — that traditional search engines, news aggregators, and online public discourse will be overwhelmed by synthetic content.

Updating foundation models with content generated after early-2023 will risk a feedback loop of ever-increasing machine-generated noise to the human-generated signal, causing AI to choke on its own outputs. As well, training models on future web content will increasingly subject them to an array of injection attacks we have yet to even imagine. Such attack strategies will no doubt be engineered with the help of the very same generative AI targeted for attack!

Reducing inputs to only sources vetted manually would be impractical and necessarily introduce strong biases, as any such vetting, by its nature, limits participation to only those predisposed (or at least willing) to abide by whatever arbitrary standards of expression have been imposed by the gatekeepers. Additionally, the immediate future of journalism, technical writing, legal analysis, and other such trusted resources will be largely automated, making them polluted datasets as well, albeit in an ostensibly benign fashion.

With traditional search engines rendered useless, next-generation search engines will instead attempt to leverage generative-AI agents, e.g., ChatGPT, to create bespoke content on the fly, downplaying links to other websites in the search results and eventually forgoing them altogether. This could allow such a search provider to effectively mount a hostile takeover of the web as an information resource and cultural medium. However, on the upside, human-moderated webrings might make a comeback. Those were fun.

.