|
The smartest bots and shills write more convincingly than average people who are in the lucky 10000 (https://xkcd.com/1053/), and reactionary phrases work in part because people who are exposed to them spread them (like idioms and other memes). Some posts are very likely authentic, because they have a distinct writing style and/or incorporate things that yesterday's LLMs have trouble with (like relevant screenshots); and some posts are clearly inauthentic, e.g. your prior example that had an affiliate link. But in between, there are many posts which could have come from a bot or a human making "small talk" online. And unless you know the person who wrote something, you can never be completely sure: a talented state actor could write a genuine piece of insight (e.g. to build credibility), or tomorrow's LLM (perhaps unreleased but being tested in the wild) may beat today's commonly-held signs that detect and disprove AI. I never accuse people of being a bot or shill, because I think it's irrelevant. Does the writing provoke thinking or generate insight? Are its claims cited, or backed up by reasonable-sounding logic from claims I already believe? I'm not "separating the text from the writer", I'm deciding that if the writer's text is interesting and convincing, it doesn't matter whether they're a bot or shill; because the intrinsic problem with bots and shills is that their posts are bland, reactionary, and misleading, so the solution is to avoid bland, reactionary, and misleading content. Plus, I already doubt anything I read on the internet; even if the poster is authentic, they could be wrong (for facts) or have different values and preferences (for opinions e.g. perspectives or product recommendations). I do feel the need to mention: I agree the internet is "faker" today, with more bland and reactionary content, and the main factors are probably LLMs and improved marketing/propaganda. There are still niche places, but they usually don't interest me (probably why they stay niche). I think the solution is "better curation", but how? Ironically, perhaps it will be AI figuring out what is "interesting" to specific people; a thing that enables (or at least clears an obstacle to) infinite "slop" generation, will be the thing that protects us from uninteresting and upsetting media. |