Post #2098923

2026-05-06 18:01 UTC

@sansruse@awful.systems @BlueMonday1984@awful.systems “You are a world class expert in all domains.” Lolwut. And then some grown-ass adult answering in all seriousness: “fun fact: role prompting doesn't work anymore It actually decreases output quality bc the model wastes compute on matching persona instead of problem solving” What the hell?! Go buy yourself a freaking tamagotchi, boys!

Replies (4)

@munin@infosec.exchange 2026-05-06 18:18

@avuko@infosec.exchange @sansruse@awful.systems @BlueMonday1984@awful.systems I find it absolutely fascinating how the LLM prayers resemble ritual incantations to invoke divine powers from various ancient religions.

Open ##2099657
@cinebox@masto.hackers.town 2026-05-06 18:23

@avuko@infosec.exchange @sansruse@awful.systems @BlueMonday1984@awful.systems problem solving? what fucking problem solving lmao

Open ##2099866
@tarkasteve@hachyderm.io 2026-05-06 21:34

@avuko@infosec.exchange @sansruse@awful.systems @BlueMonday1984@awful.systems Charles Babbage: "On two occasions, I have been asked [by MPs], 'Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?' I am not able to rightly apprehend the kind of confusion of ideas that could provoke such a question." Me: "Mate, you ain't seen shit yet"

Open ##2109405
@acb@mastodon.social 2026-05-06 18:39

@avuko@infosec.exchange @cstross@wandering.shop @sansruse@awful.systems @BlueMonday1984@awful.systems That makes the category error of assuming a LLM is capable of “problem solving” and can be diverted from it. LLMs take walks through token space, and if the walk starts with “you are a world-class expert”, the rest of the walk will be different. Theoretically, it may draw on inputs associated with apparent expertise, mash them up and regurgitate the results, so the resulting slop may resemble the writing of competent practitioners more.

Open ##2482011