I Have No Mouth, and I Must Go Against OpenAI Use Policy
– philmv (Mastodon link)
In November 2022 OpenAI revealed ChatGPT, a Large Language Model (LLM) chat interface that would answer questions, write poetry, whistle and dance, all based on data it had scraped and trained on from the internet (copyright be darned). It can be useful (it helped me write some code for this blog) and it can be silly/weird/dangerous (giving famously bad answers, of which there are too many examples to name).
But it also, more insidiously, provided a new tool to spammers, scammers and their ilk.
Futurism has one of many stories on just one tiny slice of this: Amazon Is Selling Products With AI-Generated Names Like “I Cannot Fulfill This Request It Goes Against OpenAI Use Policy”
People were already complaining about Google search results getting worse, and now it’s filling up with AI-generated junk. That same junk is also serving to help train future LLMs and you see where this is going: an internet-wide ouroboros of spammy, misleading nonsense choking out anything written by actual humans.
My observation on this, coming as we near mid-January of 2024, is already old news. I just wanted to record it here for posterity. Let’s check back and see how things have evolved by year’s end.
