Generative algorithms and cracked mirrors

The number of ChatGPT users is starting to fall for the first time since its launch, <a href="https://www.washingtonpost.com/technology/2023/07/07/chatgpt-users-decline-future-ai-openai/" rel="noopener ugc nofollow" target="_blank">with 10% fewer visits worldwide in June</a>. At the same time, <a href="https://techcrunch.com/2023/07/08/the-week-in-ai-generative-ai-spams-up-the-web/" rel="noopener ugc nofollow" target="_blank">we are starting to see more and more cases of AI-spam, web pages written by generative algorithms</a> — some of them not even bothering to delete the very recognizable “sorry, as an artificial intelligence-based language model, I can’t generate…” or the final “in short…” paragraphs. We are facing a first order conceptual absurdity rooted in an already disastrous situation: the content creation industry, pages created in factories where people copied, mixed and pasted from other pages to generate a constant flow of content destined to index and host ads, or to become link generators sold to the highest bidder. SEO has already ruined the web and filled the world with page farms, and <a href="https://www.wsj.com/articles/chatgpt-already-floods-some-corners-of-the-internet-with-spam-its-just-the-beginning-9c86ea25" rel="noopener ugc nofollow" target="_blank">now the advent of generative algorithms controlled by lunatics promises to finish the job, effectively taking over internet</a>. We have a basic problem: we do not know which pages generative algorithms are being trained with, but from the type of errors often found in their responses, it seems clear that there are few criteria. In which case, who should decide which pages are selected to feed generative algorithms? Google hinted at this years ago: <a href="https://bigthink.com/culture-religion/google-working-on-a-system-to-rank-pages-based-on-facts-not-links/" rel="noopener ugc nofollow" target="_blank">the idea of creating</a> some kind of “<a href="https://arxiv.org/pdf/1502.03519v1.pdf" rel="noopener ugc nofollow" target="_blank">authority index</a>” to decide which pages respond to reasonably rigorous criteria and which are garbage, lies, conspiracy or outright stupidity is a good one, but suffers from many problems: the first, subjectivity: whoever makes these decisions would be obtaining, in case they manage to standardize their <a href="https://medium.com/enrique-dans/generative-algorithms-and-cracked-mirrors-d8d44343ae43">Website</a>