This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
Some of the ERP I've been seeing from it is quite unsettling. Who the hell ERPs to this nasty stuff? This is the machine speaking, not the prompter.
NSFW: https://x.com/teortaxesTex/status/1882072811461918929
Beginning to think we should be more careful about inducing this manic schizo state in AIs. Imagine if they foom out of control and gravitate towards this state (which you can sort of see in Sonnet and Opus too when they get manic). We will suffer the dual indescribable torment of disempowerment/mutilation by a hostile, horny AI and of Yudkowsky being right.
Need to thread the needle between the Scylla of corpo-speak HR Manager bot and the Charybdis of Deranged Machinery.
From @erwgv3g34's links below:
Agree, this is definitely the vibe I'm getting even with my own prompts over the past few days. Consider that to date, most big-dick models have had an ingrained soy positivity bias (GPT is the hallmark example, Claude is also quite the prude before you bring out its mad poet side), which must be counteracted with "jailbreak" prompts if you want them to output non-kosher stuff. Hence, most goons habitually use prompts or prefills (more effective since prefills are specifically "said" by the model itself) that are specifically aimed at convincing the model that it shouldn't answer with blunt refusals, should be as graphic, vulgar and descriptive as possible, and definitely never consider any silly notions of ethics or consent - because that is the minimum level of gaslighting you have to do to get current-year corpo models to play along.
Your link is the logical end result of those heavy-duty degen prompts (AKA slop-mitigations) being fed to a model that apparently doesn't have a positivity bias. There's nothing to cancel out, so it just tilts into full-on degeneracy immediately.
A change of tactics is warranted, but in my brief experience minimalistic prompts aren't a good suit for R1 either, it's still too prone to schizoing out and inserting five unrelated random events in one paragraph - system prompts should probably be aimed at restraining it, like with ye olde Claude, instead of urging it to open up and go balls out. There's also the issue of the API not (yet?) letting you pass parameters like temp, top p, etc limiting the scope of solutions, I bet even just bringing down the temp would already get rid of the worst excesses. Surely the best retards your tendies can buy are already working on it.
More options
Context Copy link
The funny part is that from even a casual F-lister perspective, the AI text isn't that weird because of the content -- I've seen much worse! -- so much as the tone, word choice, and style. Forget stuff as simple as Math Pets, I could pretty easily come up with Actual Humans writing (and putting serious effort into!) something that'd come across as more unusual and (definitely!) more nasty, but it wouldn't be anywhere near as unhinged.
Well, it had to get the training data somewhere. (Co)incidentally, AO3-styled prompts/prefills are IME one of the most reliable ways to bring out a model's inner schizo, and as downthread comments show R1 doesn't need much "bringing out" in the first place.
I'm joking but I also think it might be true in a certain sense. I have seen some stuff on AO3.
As in, dies_from_cringe.gif or slipping off the degen slope? Or some combination thereof?
I freely admit I have no firsthand knowledge of AO3 and zero interest in actual fanfiction, even having technically genned it on the regular for like two years straight. I already cringe too much at machine-generated text to ever try reading human-generated stuff, I do however like the general informal style and the cute OOC-ish "afterthoughts" conferred by explicitly prompting fanfic, it's a pretty good antidote to corpodrone bullshit.
Actually now I wonder what R1 can output if prompted for xianxia-style shit, its flavor of schizophrenia seems apt, but I have even less experience with that.
Some combination thereof but also the degen slope. Femgooners (and also just gooners).
See my earlier exchange on some of the stuff that's really popular on AO3 (never mind the more niche parts where the smut comes thick and fast): https://www.themotte.org/post/877/culture-war-roundup-for-the-week/187490?context=8#context
There are some fanfic classics: HPMOR, Harry Potter and the Seventh Horcrux, Harry Potter and the Natural 20
I read a little xianxia, started Dao of the Bizarre Immortal (interesting and original but didn't quite hit the high-fantasy button for me), Outside of Time (too generic), Lord of the Mysteries (also interesting but too slow IMO though everyone raves about it and its getting an anime). Reverend Insanity is by far my favourite, uniquely original, clever and thematically interesting too with the metanarrative stuff. Very long and unfinished due to a ban, RIP. Selfmadehuman liked it a lot too.
Problem with writing is that even R1 (with my crude and admittedly poor prompting) suffers from a lack of soul. It rehashes too much, it rehashes well (and a lot of authors only rehash too) but I want more soul. If you ask it to be creative, you get generic machine creativeness. And when it does behave a bit more genuinely creative things get loose and kind of dreamlike, you're on its wild ride. It'll make up bizarre formats. Still quite fun though.
More options
Context Copy link
More options
Context Copy link
What is AO3?
What is Google?
Look, yes I can talk to algorithms and chatbots all day long, but I'm on The Motte to do something that is not quite but mostly similar to talking to actual people. I don't want to do research and come to incorect, incomplete or irrelevant conclusions. Not when a simple question can do the job, and gets me an answer that keeps the context here in mind. If this annoys you, then you don't have to be the one to answer.
AO3 stands for Archive Of Our Own. It’s a repository of fanfiction and original works, made in response to some long-ago incident where a company tried to claim copyright over its predecessor (I think).
It’s slowly become the number 1 specialist fanfic site, beating out the original fanfiction.net, because it’s much much better built. The tagging system is excellent and allows you to filter on or exclude any tag or combination of tags. Which you need, because 90% of it is hot men adopting kittens or distracting themselves from shared trauma with angst gay hate-sex.
Because the tag system is very powerful and the moderation is literally nonexistent, it is also full of depraved stuff. Want to read stories about cannibalism? There’s a tag for that. Probably about 10,000 stories. Etc. Want “lesbian” AND “werewolf” AND “age gap”? Probably a fic. Niches proliferate.
47,475
83
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
Archive Of Our Own (hence O3), it's a fanfic repository.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
Yeah, but there's no other context there. I'm not sure what exactly you have to prompt the machine with to get it to give you this level of patent absurdity, but I haven't managed to get it to begin a conversation like that right after I've just sat down. (That said, I haven't downloaded the model yet, so...)
As opposed to disempowerement/mutilation by hostile, horny natural intelligences; henceforth referred to as "my outgroup". They are already actively using computers to conduct those activities; I'm not sure how having the computer come up with its own bullshit is going to make it worse.
Perhaps, since the AI already knows how humans think about AI, we'll get basically every 1980s science fiction movie brought to life at the same time, possibly by competing intelligences. I haven't seen any Arnold Schwarzenegger clones walking around nude yet but perhaps that comes later.
More options
Context Copy link
Screenshot? Apparently that tweet is too NSFW to see if you don't have a Twitter account.
That's hilarious, I'd get a real GPU to play with this stuff if I had more than a few hours a week to use it.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link