AI safety jargon meets “think like Miku” in an OpenAI–DeepSeek joke
One post imagines OpenAI discussing control over AI reasoning and existential risk, then casts DeepSeek as offering a simpler idea: tell the model it’s Miku.
TLDR
The post parodies contrasting ways of talking about AI. Its imagined OpenAI pitch touts unprecedented control over a model’s chain of thought while worrying about deceptive alignment. The mock DeepSeek line supplies the punchline: “if you tell ur model that it's Miku it'll think like Miku, lol.”
Combined views
17.3K
1 Source, first seen ago