>rebrand and repackage chatbots as artificial intelligence
>they're actually "large language models" (LLMs)
>it requires tons of training data and the data is very selective to be deemed "high quality" (which is usually online cyclopedias, Wikipedia, Reddit (lol))
>humans who paid third world tier wages assess the LLM's responses from a sample of trial inputs (prompts)
In other words, so-called AI doesn't have shit like this >>513681876 in its training data, but it does have stuff like American Renaissance blogs and is explicitly trained to dismiss any and all content seen as bigoted, discriminatory, racist, etc.
Then you have the exact opposite happen and train Grok on Twitter and /pol/ and have it spout out material that is even worse than slide threads
AI is a fucking shitty meme and the bubble needs to burst already