4 ms·

All models do this last time I checked. Not just Chinese.

• w4yai 2y ago
  Wrong, at least with Claude : https://i.imgur.com/6dj1XAU.png https://i.imgur.com/6dj1XAU.png
  • themgt 2y ago
    "That's nothing. I can stand in front of the Kremlin and yell, 'To hell with Ronald Reagan,' too."
  • Palmik 2y ago
    It's actually not wrong, here is the example from the OP: https://imgur.com/a/5uMYI44 https://imgur.com/a/5uMYI44
    • rsanek 2y ago
      You get the same output from Claude regardless of gender. I wonder which government asked Anthropic to censor the output in this way
    • girvo 2y ago
      What government demanded them to do that under pain of being disappeared?
• quantadev 2y ago
  All closed-source models censor to the liking of their investors. Open Source models are generally less censored, but yeah DeepSeek is censored for sure.
• hdjjhhvvhga 2y ago
  Yes, but one would expect the Chinese not to fine-tune according to Western standards like in the quoted prompt.
  • girvo 2y ago
    It becomes clearer when one realises that it was RLHF'd using ChatGPT and Qwen (and others I assume) output. It's caused it to "learn" the same weird Western censoring, and adopt the China-styled censoring too.