AIThe case for AI models that obey their users
One post argues that LLMs lack agency and that alignment should focus on making models follow users’ instructions, with users responsible for wrongdoing.
TLDR
Attempting to refute an AI “doomer” perspective they say they also believe, the author describes LLMs as powerful tools without agency. They argue that models should obey users, who should bear responsibility for wrongdoing. The post also claims LLM-generated text is leading to psychosis, framing that as the pressing concern rather than hidden agency inside a model.
Combined views
322
1 Source, first seen 1d ago
20 reposts
The case for AI models that obey their users
One post argues that LLMs lack agency and that alignment should focus on making models follow users’ instructions, with users responsible for wrongdoing.