4 ms·

I think the defense would be something along the lines of "AI responses aren't representative or official statements".

kstrauser 2y ago
I believe that would be hard to defend in court. "Did someone or something inside your company say these words to the plaintiff?" "Yes." You can only disclaim so much, especially when you're making money off the output of the (person|AI).
- sodality2 2y ago
  Character.ai doesn't seem to have direct monetization mechanisms. In addition, sites like HN aren't generally held responsible for everything a user says. They could try to argue that the characters are sufficiently influenced by the user-generated prompts and user-facing conversations to be no longer their own. (Section 230)
  - kstrauser 2y ago
    But they're still acting on the company's behest. If I hire a jerk to work tech support and they insult or cause damage to my customers, I don't get to say "shrug, they don't represent my company". Of course they do. They were on my payroll. I think it'd be pretty easy to argue that the AI was performing the duties of a contractor, so the company should be responsible for its misbehavior, just as if a human contractor did it.
    - sodality2 2y ago
      Do you think that applies to open source models, or is it the act of performing inference that makes it an act the business is responsible for? ie, Meta's Llama does the same thing.
    - mewpmewp2 2y ago
      But with Character AI you are hiring a roleplay service which can be open ended in terms of what you are looking for. If you are looking to roleplay with a jerk, why shouldn't you be able to do that, and in such case why should the company be held liable?
  - mewpmewp2 2y ago
    In any case I think society and building of society should be directed in such a way that we don't have to censor the models or baby the models, but rather educate people on what these things really are, and what makes them produce which content, for which reasons. I don't want to live in a society where we have to helicopter everyone around in fear of one single misinterpreted response by the LLM.

I believe that would be hard to defend in court. "Did someone or something inside your company say these words to the plaintiff?" "Yes." You can only disclaim so much, especially when you're making money off the output of the (person|AI).

sodality2 2y ago
Character.ai doesn't seem to have direct monetization mechanisms. In addition, sites like HN aren't generally held responsible for everything a user says. They could try to argue that the characters are sufficiently influenced by the user-generated prompts and user-facing conversations to be no longer their own. (Section 230)
- kstrauser 2y ago
  But they're still acting on the company's behest. If I hire a jerk to work tech support and they insult or cause damage to my customers, I don't get to say "shrug, they don't represent my company". Of course they do. They were on my payroll. I think it'd be pretty easy to argue that the AI was performing the duties of a contractor, so the company should be responsible for its misbehavior, just as if a human contractor did it.
  - sodality2 2y ago
    Do you think that applies to open source models, or is it the act of performing inference that makes it an act the business is responsible for? ie, Meta's Llama does the same thing.
  - mewpmewp2 2y ago
    But with Character AI you are hiring a roleplay service which can be open ended in terms of what you are looking for. If you are looking to roleplay with a jerk, why shouldn't you be able to do that, and in such case why should the company be held liable?
- mewpmewp2 2y ago
  In any case I think society and building of society should be directed in such a way that we don't have to censor the models or baby the models, but rather educate people on what these things really are, and what makes them produce which content, for which reasons. I don't want to live in a society where we have to helicopter everyone around in fear of one single misinterpreted response by the LLM.