For Section 230 purposes, is AI generated text third party or first party content? If a site sets up basically unsupervised or algorithmically supervised routine publication of #ChatGPT content, is it under current US law liable for what the robot says? Would OpenAI be? Would anyone at all be? (inspired by @emilymbender https://dair-community.social/@emilymbender/109716249225156006, though she presents a more traditionally publication-like, so arguably more likely liable, case.) #Section230
@interfluidity Usually the AI content only gets shown to a single user in the course of interacting with the bot, so it's a bit orthogonal to the usual Section 230 debates about giving content a widespread platform
@yarrriv I'm not asking about the you-and-ChatGPT have a talk case, but hypothetically if a site routinely solicits and publishes for a wide audience (like CNET is doing, albeit with human editors, in the piece @emilymbender highlights) content from LLMs, who if anyone is liable (especially if there are no human editors in the loop).
@interfluidity There's always a human at some level of decision-making there, though, even if they don't approve each singular piece.
You can imagine a parallel where a newspaper had an inbox anyone could write to, and featured randomly selected emails on its front page without examining the content. Would they still have Section 230 protections in that case? I would guess not but IANAL
@yarrriv A newspaper would not, but a website that features randomly or algorithmically or idiosyncratically curated "letters" does. That's precisely the effect of #Section230, for better or for worse. Websites are treated quite differently than offline publications that include "third party content".
@interfluidity the one thought that occurs to me is that perhaps we should bring in the sort of concepts they use in adjudicating negligence, where you can be at fault if you didn't make some reasonable efforts to make your bot behave
@yarrriv @interfluidity That would be a very difficult line for me to draw, and I'm a career software engineer with an interest in law and knowledge of the subject material. I'd be terrified to find out where legislators try to draw it.
@LouisIngenthron @interfluidity I think standards of reasonableness are usually figured out in courts rather than by the legislature, so they're not clear cut lines
@yarrriv Which sucks, because judges are arguably worse than legislators about technology.
Determining intent from code is difficult in the best of circumstances. When talking about neural net algorithms, it becomes essentially impossible.
It's also a fledgling technology with few standards or rules to measure against. Who would even have the knowledge and capability to write such standards without risking serious chilling of innovative potential?
Without being able to determine intent, nor having a standard to measure up to, how can we even think about being able to determine negligence?
@LouisIngenthron sure, but the other option here becomes making "the AI did it" a get out of jail free card, which doesn't seem better
