Comment by 2-3-7-43-1807
1 year ago
the attempt to elicit slurs is just a way to channel and vent frustration about much more complex and practically relevant restrictions caused by censoring in llms. it's just the most simple and primitive illustration of a larger problem with ai.
Unless you can set out some of these ‘complex and practically relevant restrictions’ that just sounds like a high-falluting attempt to justify trying to elicit slurs.
huh? what is there even to justify about that? are you worried the llm gets traumatized?
This is a beautiful explanation, it's tempting to add up another take here: these companies exploring AI safety are really just selling their anthropomorphised machines and making good money. The intelligence they sell is so intellectual it needs a word-blacklist in order to look safe for legislation.