Raw LLM Responses

These bots are using all the lessons of the spam trade, make shit full of punctuation, grammar, and spelling errors. Have gaps in your sentence structure etc. It's gotten to the point where you can give a chatbot a corpus (whole bunch of text written by someone) and it will start spewing academic papers that pass initial scrutiny (academic papers mostly about nothing... but hey.) Folks attention spans are short enough, and the initial cost is so low to set something like this up -- expect to see tons of it. Feed in the algorithmic picture generation (more machine learning) and you can generate random pictures of the same fictional person (in different poses/clothes/scenery what have you) and you can generate \*as much\* content as the average user. Finding this stuff for Facebook and Twitter is getting much harder, figuring it out as your average Joe will eventually take much more than critical thinking skills, it'll take open source intelligence research skills... I hate to advocate something like Estonia's Internet ID (smartcard based identity) but... without some hard token issued by a trusted body -- that is tied to your identity, I don't get how this is fixable any time soon. And it's only going to get harder as the bots get smarter.