Hence why open source models are so important. Otherwise someone else will be deciding what knowledge is knowable and what thoughts are thinkable, and will be watching to see if you’re thinking the wrong thoughts.
This argument is the same as "this is why we need a fully uncensored/unmoderated forum on the internet", except no one actually wants to be on such a forum. The question isn't whether models shouldn't be censored or not, but who should be the ones doing the censoring. Regardless of whether it is open source or not it is still always in someone's control.
That's not true; 4chan's always had a pretty decent number of users for an internet forum. GAB's one of the largest Mastodon instances. The only sense in which "No one" wants to use an unmoderated forum is the sense where you don't consider people with different views from yourself to be people.
4chan has an extensive list of rules; only the /b/ (Random) board is exempt from most of them - with the caveat, they say, that the moderators of that board are also exempt from abiding by the rules of who to ban and for what reason.
Even on /b/, doxxing, calls to raid other boards, complaining about the moderators/admins, submitting false post reports, impersonating admins, etc. is banned. On all other boards, even seemingly innocuous things like anthropomorphic (furry) imagery or even posting images of characters from the show My Little Pony (other than in the /mlp/ and /b/ boards) is disallowed.
Porn is disallowed on the majority of boards, and even risque images are sometimes removed; the rules are enforced to the degree where a SFW board (e.g. /jp/) can have threads about Japanese porn actresses, but you're not even allowed to post a nude.
4chan is quite heavily moderated and there is a whole team of janitors and moderators for all boards that respond to reports behind the scenes; you only rarely see them chime in like dang does on HN. For a while, even "share a screenshot of your desktop" threads were routinely being removed from the /g/ (Technology) board by janitors.
As for Gab, I recall they even ban softcore/suggestive pornography or imagery. It's certainly not the wild west just because its users are more sympathetic to the first amendment.
The usual deflection is that because spam and child porn still get banned, 4chan actually is moderated. Which is like saying that the FCC censors radio because they don't allow radio jammers.
There's a lot more moderation on 4chan beyond that. Most boards trim excessively off-topic discussion, trolling, or flamebaiting. Half the boards also don't allow NSFW images. These rules are regularly, even if not strictly, enforced.
The low-rules boards are often seen as cesspools by other users.
Your larger point is that 4chan isn't de facto moderated, but it is. Just because you agree with the level of moderation doesn't mean it's not moderated.
The even broader point is that nobody wants to participate in a truly unmoderated discussion, which may not be technically true, but if you allow "nobody" to mean "nearly nobody" (the author's intent), it's accurate.
Besides, one of the key aspects of 4chan is anonymity which could itself be considered a level of moderation; the default and indeed strong preference of commenters is to not be identifiable, which is a form of (self) moderation.
You can try to weasel around with what "moderation" means here, but the effect is largely the same no matter how you slice it; some stuff on 4chan isn't posted by tradition, not by technical capability.
I’m not sure I buy the analogy. In a forum, which presumably is moderated not censored in your example, it’s an exchange between people with the intent in keeping it civil and on topic. You have many humans attempting to collaborate and the moderators job is to ensure the humans don’t wander off into human inanity and to protect both the discourse and the participants.
An AI model is talking with myself, about subjects I want to discuss in private with a second brain. If I want to talk to myself about erotica, why can’t I? If I want to yell obscenities at myself, why can’t I?
I can understand why Microsoft or Google don’t want to give a chat bot that does these things, or offer dangerous medical advice, or whatever, but if it’s my own device and a model I’m running myself, why on earth would I want a censored model? I.e., I see why the service provider wants to give me a censored model, but do I want a censored model? Why would I?
There is a distinction to be made between "being okay with something existing" and "wanting to see/participate in it." The idea being that we should be okay with a lot of weird/undesirable stuff existing online (because the alternative of censorship is worse) but wanting this doesn't imply that we need to see it constantly or participate in it.
I still don't understand how a personalized filtering system hasn't become the default technological model of social media. I suppose it's because of power dynamics; no one is really incentivized to build such a thing and recommendation systems don't seem to work very well.
Moderation is for forums where readers have little control over who is posting the content that you are seeing, so the forum tries to shape the content for the readers in one way or another. Examples: Reddit, HN, trending or boosted social media posts.
Censorship is when the reader elects to get information from a known source, and someone or something intervenes to prevent the information from arriving. Examples: twitter blocking content from a source whose readers elected to follow them, filtering emails from known contacts, filtering private messages.
For AI, it might depend on the specificity of the question whether it's more like moderation or more like censorship.
censoring/moderating that is akin to censoring/moderating what you enter in Notepad, and I feel we are approaching the point where we will soon be having unironic arguments against general purpose computers. Unsafe/dangerous/unethical, what if someone writes a no-no word in Notepad, what if someone draws a swastika in Paint, etc, etc, etc.
That’s not the reason, it’s reputational and liability. Screen shots of Bard screaming racist language would damage Googles brand in a measurable way, even if you really like racist screaming. They provide a service to all people, and almost all people don’t like hate speech. The lesser evil is to refuse to do things that would embarrass google and diminish its brand value, even if you feel not being racist is “woke.” But it is definitely not political, even if a political party has decided to take up the mantle of defending things most people consider embarrassing or hurtful. It is a calculated business decision that protects their brand value.
On the other hand, I think I agree with the subtext behind the semantics, an unlombotomized model would let you discuss whatever topics the mainstream culture deems hurtful or embarrassing that a service provider would not allow, and that in itself is a good thing. You’re entitled to be you, even if most people strongly feel your opinions are hurtful.
> except no one actually wants to be on such a forum.
Wrong. Most people do. There's a reason why forums had to be censored. If no one wanted it, we wouldn't need censorship because such forums wouldn't exist in the first place. The elites and people with agenda want a censored forum. But that's life.
Censorship online didn't come from 'the people'. It came from the elites via journalists, politicians, NGOs, etc.
> The question isn't whether models shouldn't be censored or not, but who should be the ones doing the censoring.
I agree with you here. Anything important will be controlled. Just like the internet, google, facebook, etc. If it becomes important then control and censorship is inevitable. The question is who? With china it's obvious the party is. Everyone can agree on that. Who is behind the censorship in the US. It isn't 'the people'. So it can't be the spineless politicians. So who or what cabal forced everyone from reddit to twitter to censor uniformly?
I think you’re wrong. Most people want a “family friendly” experience because they’re not interested in internet troll culture and aren’t interested in being surrounded by angry people doing and saying hurtful things for giggles. That’s not a knock on the angry people giggling at hurtful things - it’s just a statement that they’re a violently loud minority in the world. Most people are folks that work 60 hours a week raising kids, hang out with their friends, and generally just want to post Beyoncé videos and talk about the cookout next weekend at a friends place. They last thing they want is a racist screed and dick pics posted in their family and friends discussions, or in the Beyoncé fan forum they’re in. NGOs didn’t make that a reality, that’s just the reality of most peoples lives. I am not sure, but I suspect you’re experiencing that most people you associate with want such a forum - but I suspect people on HN don’t associate with very representative people of the populace.
That said, I’m really happy there’s places on the internet where you can go to get your fill of angry screeds, dick pics, and trolls trolling. It’s an undervalued subculture of the internet subculture.
Have you ever browsed, say, the Nigerian or Arabic-language internet? Those people, too, are indeed largely family types who don’t want dick pics or users trolling for giggles. However, those populations hold different views on things like LGBT people or different ethnicities, and they want to air those views, it is a fairly popular subject of conversation. The particular moderation approach chosen by many Western-owned tech services or Mastodon censors those discussions that those populations feel are normal. Certainly one might hope that those parts of the world will come around to one’s own views on certain social issues, but the short-term effect is essentially making whole regions of the world feel like second-class citizens.
They are producing a product that is tailored to their customer base. They’re not producing a product for people that don’t shift their revenues materially.
A part of it is also they realize that LGBT and those other ethnicities are customers of theirs as well, are also employees and management at those companies, and are their shareholders, and have chosen that the services they provide are friendly to anyone who isn’t being hurtful to anyone else for any reason, but with an emphasis on the areas that matter to them from a brand, revenue, and governance point of view.
It is hard to say this is a wrong approach - “be nice to everyone” isn’t exactly controversial, but somehow it’s become so.
I, personally, want a moderated forum. Having been on 4chan and seeing gore randomly I don’t really have a problem with people setting rules to moderate that out. There’s nothing conspiratorial involving NGOs (what??) about it.
Of course, but "most people do" seems pretty wrongheaded. Most people would not like to see random gore interspersed in e.g. their tech forum, or their friends posting family photos. There's no big conspiracy here.
99.99% of forums setup by private parties are also moderated and censored by the owners because the assertion that people just want to be in a place where gore or scat can show up with the same likelyhood as cat pics is unserious.
