Meat-based LLM proxies
11 points by ashwinsundar
11 points by ashwinsundar
I see this most often on Linkedin. People will feed AI the posts of others, generate and post a comment on it, and then receive genuine responses from unwitting posters.
It is disturbing to me: they are acting as a meat proxy for the AI in order to build professional and personal relationships.
I mean, LinkedIn was full of slop even before GenAI
I wonder to which degree LLM output is optimized to appeal to the same people as LinkedIn. It’s striking how they both are producing the same kind of insipid and obsequious prose, but is one the product of the other or are they both trying to attract and feed off the same people and this is the inevitable optimum?
LinkedIn as the original crab shape in the evolutionary struggle to appeal to investors / ”business leaders” and attract investments.
I think LLMs simply made slop at scale possible. I don’t use LinkedIn (I tried once, their insistence on sending me ten billion emails led me to delete my account very quickly), but I imagine it got way worse once LLM driven industrialized slopping became available to the masses
The only thing I can say about LinkedIn is to never list your current job there! I did find the article interesting in the way that it reflected how AI's training models really do reflect their overall goals... whether it's to be polite or what not. I do find with clever prompting however, that you can certainly make some very interesting writing pieces! It's very cathartic if one is angry. Something that that you never send but say. Here is an example prompt: I would like you to write something that questions their parentage and uses at least four versions of the verb to reproduce, as an adverb, adjective, and a noun. Do this in the style of a drunken lord gentleman from the nineteenth century.
It makes for some interesting fun! Not sure how much has had to do with the original, but it is kind of fun to play around with these things sometimes.