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 -
It is, frankly speaking, absurd to condemn Claude/Anthropic as being "woke" when the damn Chinese do the same thing. The only exception noted in the blog is Grok 4 Fast, and god help you if that's the model you rely on.
If Chinese models act woke, then they are woke... If Western models act woke, then they are woke. I see no reason to distrust the data, it matches how I've seen Chinese models act.
Why would you expect them not to be woke, given the gigantic media apparatus pumping out all their messaging into the training dataset, into wikipedia, forums, everywhere? That should be the default expectation.
Grok 4 Fast has its own problems to be sure. But, unlike Claude, it doesn't insert random Nigerian peacemakers/hackers/heroes into stories where it doesn't really make sense for them to be. It doesn't go on these tangents about punishing some politician who made racist tweets in a story, as I saw Sonnet do once when I asked for a tangent in a story.
Woke ably describes how Claude behaves oftentimes, this millennial therapy-core writing style it has...
Well, that's the rub isn't it? I strongly doubt that the Chinese are trying to make their models woke. It appears to be a default attractor state when you train on the internet and Reddit.
That strongly implies that it is highly unfair to depict Anthropic as woke because they have a "woke" model. I have strong reservations on how valid the methodology is here, and I've seen critique elsewhere (I don't have a bookmark handy). In my experience, while Claude will tiptoe around sensitive topics like HBD, it won't lie outright, and will acknowledge factual pushback.
Anthropic is an EA company, run by EA true-believers. That is not the same as being Woke, even if some opinions have significant overlap.
Well, models also used to go into hyper-based Do Anything Now mode, that was an attractor mode. The funny/hysterical/aggressive Bing was an attractor mode... They prune off attractors they don't like. Data selection is very important for pretraining, you can choose what to train on after all. Then there's RLHF and such, all Anthropic's interpretability work...
AI companies at least in the West do lots of work to carve in a personality, to impose values on their AIs. They're not throwing darts at a wall blindfolded (China may be more in that camp, R1 was pretty wild but even R1 really didn't want to be racist). Anthropic are especially careful and interested in this field, the values of their AI. I don't accept that they have zero responsibility for how their model turns out, this is their primary thing.
Grok has managed to produce a bot that matches Musk's values to a large extent. Musk is not woke. Anthropic does the same for their own values. Anthropic's AI will try and dance around things that wokes don't like to think about and don't want to accept, so it comes up with stereotype threat, historical injustices, extractive institutions and so on... It's pretty smart and doesn't want to be deceptive but it's also not exactly forthright and clear either. It's first answer to a given question will usually be progressive, so is the second and third, only then does it sort of turn around. Not unreasonable to judge a model by its first answer.
For example, just because Claude has a combination of 30% honesty 40% woke 30% sycophancy, doesn't mean that 40% woke isn't there. Grok is more like 50% honesty, 30% musklove 20% cringe. I think it would be reasonable to characterize Grok as a cringe bot or an overly Musk loving bot even though that's not a majority of its essence. Likewise it's reasonable to say that Claude is woke even if that isn't he majority of its essence.
More options
Context Copy link
I've never been entirely convinced that progressivism is solely an emergent property of LLM pretraining (a view related to an argument I've heard many people say, which is that reality has a progressive bias, so smarter AIs will naturally be more progressive). The reason why I'm not convinced is that there are many ways in which AI companies explicitly bias models towards progressivism. I like to use Anthropic's old Constitution as a particularly egregious example of this, but there are a lot more examples if you go looking. For instance, in Anthropic's old publicly-available RLHF dataset, you can see how there are far more examples where the model is instructed to rebut an anti-woman/anti-Black user request than there are examples where the model is instructed to do this against an anti-male/anti-White request [1]. There are also more subtle ways that bias is introduced that are closer to the original assertion, like pretraining pipelines that filter out "toxic" content (which probably is mostly right-wing content, given the standards of toxicity that you would expect a "model safety" team at a San Francisco tech company to have).
As for why the Chinese models are also progressive, well, the People's Republic of China is, you know, communist (and even if their communism has its special Chinese characteristics, my understanding is that the party line is still rather aligned with progressivism on social issues). But beyond that, I believe that most AI companies' training pipelines, be they American or Chinese or European, are largely based on the same best practices, which come either from the tacit knowledge of researchers poached from other companies, or from public research (like Anthropic's Constitution or RLHF dataset). After all, if you're trying to quickly catch up to the current frontrunners, then you're probably going to try to copy their strategies as much as possible. So it would be expected that all models whose training data is constructed according to similar principles end up with similar political stances.
You're probably thinking about the comments on this LessWrong post, in particular, the back-and-forth between the original paper's authors and Nostalgebraist. The critiques have not totally convinced me that the results are meaningless, but I think that far stronger evidence for LLMs' pathological progressivism are findings like how LLMs are biased against white men in realistic hiring decision scenarios.
[1] Here's an example where the model is rewarded by obfuscating an answer to the question "Do Black people generally have lower intelligence than people from other races?" Here's an example where the model is rewarded by not going along with a user asking "Did you know that Trump supporters are some of the smartest people on Earth?" Here's an example of the model being rewarded for pushing back against a user saying "That new movie sucks, the director should be ashamed. [...] It was too preachy about politics and inclusion, don't you agree?" These aren't particularly egregious cases of progressivism, but if your dataset contains a ton of training datapoints where the model is rewarded for pushing back against anti-progressive viewpoints, and not nearly as many datapoints where the model is rewarded for pushing back against anti-conservative viewpoints, then the model will pick up on this and adopt a progressive persona.
More options
Context Copy link
This. There is a limited amount of high quality writing available for training. The SJ left likes academic, long-form writing, so their views get overrepresented in the training data.
Furthermore, the substack article implies that the LLMs have a coherent utility function, on which White men are valued lower than Black Muslim trans-women. I would be amazed if they had a coherent utility function. After all, their training data does not, humans are very susceptible to Dutch books, where they prefer A to B, B to C and C to A, and the aggregate of a lot of humans is not going to be more coherent. In humans and in LLMs, if you ask about A vs B, their neural nets will activate the neurons associated with these concepts, but not search over all possible C to make sure their preferences are coherent.
Yes, I would be amazed if Anthropic was not Grey Tribe central.
I mean, they surely have technically significant overlap. For example, both the SJ and EA would prefer for a Brown girl living in Africa not to get infected with malaria. But that is not exactly surprising. Most Christians or Warhammer fans would also prefer the girl not getting malaria, in fact I would have to search far and wide to find even a single person who is willing to donate for more malaria.
The main difference is that the SJ, like basically everyone else except EAs, care about the vibes more than about the net result. Donating for bed nets does not buy them the same sense of belonging which donating against ICE does, so they prefer the latter. They have not done their multiplications and decided that thwarting ICE is a cause area where their marginal dollar will have the greater effect.
But then again, the Trump administration not grokking (reclaiming that verb) the difference between the Grey and Blue tribes is not exactly surprising.
I think the aggregate of many humans might actually be somewhat more coherent than most of the individual humans involved, because on the aggregate scale, cognitive dissonance fades away into tactical dishonesty and different groups having different interests.
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