Spoken like someone who hasn’t used them. They can be wrong and require care and thought when being used but calling them “bullshit machines” is just lazy and inaccurate.
I cannot recommend this paper enough.
LLMs produce a large amount of bullshit. Also, it’s been clarified that they can never be made to stop the bullshit and only produce truthful output - because the concept of truthfulness is not a part of the way they work.
https://arxiv.org/abs/2401.11817https://arxiv.org/html/2409.05746v1
Today, when people talk about “AI” (a term with no clearly defined meaning, unfortunately), most of them mean LLMs most of the time. But when discussing “AI”, (or making memes about it) we would ideally specify that explicitly…
Spoken like someone who hasn’t used them. They can be wrong and require care and thought when being used but calling them “bullshit machines” is just lazy and inaccurate.
https://www.researchgate.net/publication/381278855_ChatGPT_is_bullshit
I cannot recommend this paper enough. LLMs produce a large amount of bullshit. Also, it’s been clarified that they can never be made to stop the bullshit and only produce truthful output - because the concept of truthfulness is not a part of the way they work. https://arxiv.org/abs/2401.11817 https://arxiv.org/html/2409.05746v1
Today, when people talk about “AI” (a term with no clearly defined meaning, unfortunately), most of them mean LLMs most of the time. But when discussing “AI”, (or making memes about it) we would ideally specify that explicitly…