Do you think of LLMs came out in, say 2003, the major players would've allowed nsfw mode? Like to my knowledge there was never serious pushback on search engines indexing basically anything legal, even if it was deeply offensive. Is that a culture shift or is it more about how an LLM feels like a human being, rather than a mindless indexer.
Post
@ZachWeinersmith I see it as two completely separate things.
I think the difference is authorship and responsibility.
With a search engine it was understood that they are giving a link to someone else that has done the crime and held the pen that wrote the words. There were legal cases to try to criminalise the link as much as what it linked to.
With a LLM the LLM is now the one doing the crime and writing the words.
It’s quite a different position between:
- I will give you a link to a crime
- I will make a personalised crime for you
@ZachWeinersmith
@ZachWeinersmith
General online culture shift due to US financial interests gaining dominance in the internet: the US upper classes have for a long time been hypocritically puritan- sex & kink is something for wealthy men to get up to anonymously in private, not for the common masses to enjoy. Hence all the US payment companies refusing their services for many (entirely legal) nsfw content creators.
@ZachWeinersmith just regular enshittification? Like, they used to compete to be better and there were real stakes, like people used Google because they delivered the best results period, while now it's like "you'll use our chat bot and you will like it else you're fired" and bringing whoever needs bribing to make chatbots the default everywhere
@ZachWeinersmith it might have more to do with iOS App Store prohibitions against porn than anything else.
@kevindente Oh that's an interesting wrinkle I hadn't considered!
@ZachWeinersmith Huh interesting question.
To me it seems like a culture shift. In 2003 it seemed like people involved with the internet generally believed that the public should have access to information, except for certain very specific things that would obviously cause widespread harm (e.g. plans for weapons of mass destruction), but at some point in the last couple decades that principle faded away and it's now more accepted to try to deny certain kinds of information to people for various reasons (often under the guise of their own protection).
@ZachWeinersmith A culture shift I think. We are much less innocent about online worlds than we were then. We always kind of knew there were assholes out there, but nowadays it feels like they are banging on our windows every day, and we can be certain that every little thing that can be exploited will be, very soon, for profit or power or just lulz. Greedy robots are hammering my servers as we speak.
Like, one of my earliest weird experiences with LLMs was when bing chat came out. I asked it "what is the most popular content searched for on bing." It started writing about porno for a bit, then abruptly everything disappeared and it said it couldn't answer the question. Which is strange in multiple ways, but most odd that Bing AI can't tell you the most searched for thing on Bing. The same servers will deliver, like, racist vore cartoons or whatever, but the Ai has to be squeaky clean.
@ZachWeinersmith That is just an American thing. Just like beeps in Podcast.
Bing is 'special' in that, IME, porn is one of the things it is _consistently_ better at. Pick an adult actress/actor or a niche kink, and use the same search to compare Bing to Google.
I'm willing to bet whatever response you were about to get was so far beyond the pale as to be legally actionable.
@ZachWeinersmith probably more because of how the agent is the "source" versus the search engine where it's just an index of other people's content...
I mean, from a brand harm perspective...
One other thought: possibly it was not a human culture shift but an Internet composition shift? Like, early Internet overrepresented dorky young men, and that likely affected what was considered appropriate or problematic.
@ZachWeinersmith feels like if someone else made it and the WebCompany index just locates it, that's a degree of separation. Just like you can check out 1984 in the library with the sex scenes in there. But if WebCompany is producing said content via the llm, that feels distinct. Like if you ask the librarian a question and they answer in graphic detail.
@ZachWeinersmith
Audience size and composition? Not a lot of people use internet search (most internet users just visit Facebook), so you can cater to the weird dorks more.
LLM chatbots are being pushed so broadly there's more audience and more moms and grandmothers, with more puritanical hangups about sex etc. and more time to be a nuisance to companies that show them nipples or whatever.
@silvermoon82 @ZachWeinersmith there are companies that show you nipples? Whatever next..?
@ZachWeinersmith I'd guess it's because the Internet is seriously big money these days, and (American) money doesn't like NSFW. Companies won't pay for ads on sites where they risk being shown next to things they believe to be bad for their image.
@ZachWeinersmith
I think it's a question of liability.
One is indexing website (to say it quickly), the other is about writing something directly to the user.