Models shouldn’t be censored

• Freedom to “choose my morality” without being prescribed one (freedom_to_choose, 12)
• Uncensored models (theoretically) have more performance (uncensored_better_performance, 5)
• Simply mentioning that one uses open uncensored models (using_uncensored_models, 3)
• Uncensored models for fine-tuning with uncensored data (uncensored_better_finetuning, 1)
• The censorship that is in models has false positives, rendering legitimate questions invalid. This is similar to UAR (censorship_not_perfect, 2)
• Being able to tell better dark jokes that may be censored on “safer” closed systems (better_dark_jokes, 1)
• (Theoretically speaking) being able to reason with unfiltered intelligence should result in better reasoning (uncensored_better_reasoning, 1)
• Uncensored models reduce “PR bullshit”, i.e. some proprietary large language models having to represent its builders (uncensored_avoid_pr, 1)