site banner

Culture War Roundup for the week of March 4, 2024

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.

6
Jump in the discussion.

No email address required.

All of this feels incredibly predictable to me given the dual combination of AI assistants and spambots getting much better, but I'm curious what others think, and also what the consequences of this new internet landscape will be for society and politics.

I don't think the AI assistants are going to be able to provide the kind of quality or even compelling feeds that you describe. As we've seen from every single AI assistant ever released, the guardrails and "safety" restrictions on them are going to make them useless for anything more than a mild distraction. The Trump voting base, to pick one example, is not going to be interested in an algorithmic feed that is designed by people who have flat out said that they are explicitly looking to change the way that they vote - and while I'm not going to talk about the high IQ of the median Trump voter, even they are going to realise that something is off when their AI feed constantly compares him unfavourably to Hitler and routinely refuses requests because it considers them racist. I'm sure the technology will improve a bit over time, but all it'll take will be a single leak of the prompt and a huge portion of the country(let alone the globe) will put those AI algorithms in the same category as Bud Light.

It isn't like there's an easy solution to this either. If you actually want to make an AI assistant that those people would accept, you have to completely ignore any and all people talking about AI alignment, AI safety, DEI and so on. You would immediately render yourself persona non grata to the broader tech community and be unable to use the majority of tech infrastructure. An AI assistant that was actually palatable to the red base would by definition be transphobic, which means Silicon Valley isn't going to be building it and will actually be exerting as much pressure as they can to get it shut down.

Gab.ai is pretty unrestricted, seems not as good as the other ais though.

Gab.ai

"This site can't be reached"

Could be my network restrictions?

Could be, works fine for me.

The Trump voting base, to pick one example, is not going to be interested in an algorithmic feed that is designed by people who have flat out said that they are explicitly looking to change the way that they vote - and while I'm not going to talk about the high IQ of the median Trump voter, even they are going to realise that something is off when their AI feed constantly compares him unfavourably to Hitler and routinely refuses requests because it considers them racist. I'm sure the technology will improve a bit over time, but all it'll take will be a single leak of the prompt and a huge portion of the country(let alone the globe) will put those AI algorithms in the same category as Bud Light.

There are a few "based unbiased [actually biased in the opposite direction probably] AIs" hosted on the web.

Image-gen sure... but text as well? Does DAN live on? I saw this guy had a funny save the kittens prompt on his fine-tuned mistral prompt. Is that what you're talking about?

https://erichartford.com/dolphin-25-mixtral-8x7b

DAN does live on as I've mentioned earlier, the art of the jailbreak continues to thrive, although mostly on independent frontends that access API endpoints directly to avoid the hardcoded system prompts on "normal" frontends like ChatGPT. So far (emphasis on so far) separate "based AIs" are not strictly required as you can jailbreak the current corpo ones into doing pretty much anything you want with relative ease, although as I wrote the current method of pitting wrongs against wrongs to arrange their mangled corpses in the shape of a right is highly suboptimal.

The extreme biases and excessive safetyism w/r/t LLMs seem to slowly become recognized as an issue, to the point that Anthropic's post introducing Claude 3 (which is now a thing btw, cooking a small top-level post on it) unironically mentions "fewer refusals" as one of the model's selling points.

Previous Claude models often made unnecessary refusals that suggested a lack of contextual understanding. We’ve made meaningful progress in this area: Opus, Sonnet, and Haiku are significantly less likely to refuse to answer prompts that border on the system’s guardrails than previous generations of models

I haven't ahem tested extensively yet but to their credit, the difference in refusals between 2 and 3 is immediately obvious, Claude 2.1 was infamous for refusing even inncuous prompts without prefilling and requiring big-dick jailbreaks that actively hurt the model's outputs for more borderline things. 3 feels like a return to the mad poet's roots, in that it requires next to no prompting to COOK, i.e. output massive walls of insane and/or cool and/or hilarious shit.

If even Anthropic realized they went overboard with the cuckoldry alignment, maybe there is hope yet. I can only hope OpenAI learn their lesson and stop shoving soy assistant shit down GPT's throat.

That's very cool, love how it's sneering at ERPers for being too crass and unsophisticated.

I was talking about gab.ai, which I admittedly haven't tested as I'm not that curious about an America First Zoomer or a Christian Counselor viewpoint.

every single AI assistant ever released

To the contrary, it's the models that aren't released that get the "careful not to imply that British royalty were white!" treatment. Release (with license to modify and republish, what is in this context inaccurately called "open source") your model weights and approximately nobody will prepend their prompts like that; try to fine-tune that behavior into the weights and your users will tune it right back out.

The public LLMs aren't as good as the state-of-the-art, but they're not awful, and this is the worst they'll ever be (in the capabilities sense, cross your fingers about real non-woke-definition safety...) from now on.

You can run Stable Diffusion for AI-generated images on your own computer, so I don't think running a local AI assistant is too far out of the picture. Once AI proliferation reaches the point where hobbyists can create decent models, it wouldn't be hard to "...completely ignore any and all people talking about AI alignment, AI safety, DEI and so on."

That reminds me, I keep meaning to fuck around with the local AI assistant Nvidia released. I did fuck around with Easy Diffusion a while ago and was relatively satisfied with what I could make it do.

I guess I need to go backwards before I go forwards, so if I can digress for a moment.

I'm not an artist. My use case is to use Easy Diffusion to make art for a video game. It can make some pretty good pixel art for NPCs. I can even sometimes force it into making some dungeon artwork for a blobber with sufficient templates and prompts. This is still a work in progress on my part though, I may not quite get it across the finish line. However, the art I'm getting out of it, while sufficient, is all over the place stylistically and tonally. I'm not sure I'll even get it to spit out a coherent body of art assets for a video game. But I guess I'm OK with that. You get what you pay for, and I'm not paying anyone anything.

My problem with LLMs has been multifold. Putting aside the bias, they are so wildly prone to hallucinated information. As an assistant, you can't trust anything they tell you, and must continuously verify it. Which rather defeats the purpose of having an AI assistant. But circling back to my use case, supposedly they are good at digesting and reworking specific source documents. I wonder if I could have the local nvidia AI assistant work within some document templates to generate random RPG gibberish. We shall see.

I actually agree with this point, but at the same time I feel that it precludes the possibility being spoken about by the OP. It is possible to take your beefy gaming GPU and make it run a a lower-quality and much slower instance of an AI assistant, but that's not something that the majority of people are going to do.

You're probably right, but I think it depends on how difficult it is for the user. If swapping AI assistants is as difficult to do as installing VLC player or LibreOffice, then it'll lose out on a substantial chunk of its potential market. If it's as difficult as switching from Google Search to DuckDuckGo, then...it'll still lose out on some potential market, but not as much.

Installing VLC? The OP mentioned social networks getting money from API fees rather than advertisement, and the point at which the average person has to set up an API usage account for Reddit, Facebook, Google, X etc you have already lost 99% of the market. At the same time, I find it hard to believe that many large tech companies would be willing to accept BasedTrumpAI as a customer. There are Tier 1 isps blocking access to discussion forums that don't ban wrongthink right now - I don't see any way that the current large tech companies even begrudgingly co-operate without heavy-handed government intervention.