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 -
A question I've had for a while is that, well, given that LLM text written for humans is so uniformly bad, why do we expect anything else it produces to be worthwhile?
The Motte is a relatively pro-tech space. I expect people here to be more sympathetic to LLMs than the median person. The Motte is, after all, several steps removed a spin-off from SSC and Less Wrong before that, both sources that have swallowed LLM/AI technology pretty enthusiastically. Nonetheless here, when it comes to posting, there seems to be a pretty strong consensus that LLM-posting is bad, and the text that they produce is worthless slop.
Shouldn't that incline us to skepticism regarding everything else boosters and hype men say about these bots? If you read text from an LLM, it's garbage. If you talk to an LLM for an hour, the overwhelming impression you'll come away with is how dumb and uninteresting they are. This is supposed to be the rapidly-improving super-intelligence that will obsolete us? Really?
LLM text written for humans can be very useful. For quick research, short of having a team of experts on retainer who are obliged to answer you within a minute or two if you text them, the bots are obviously the answer. If I am debugging a specific Linux problem, I am not going to care that Scott could have presented the answer in a better writing style.
Here is a CW analogy for you. Guns are undeniably useful. Cheap. Scale well. Incapacitate physically stronger opponents without ever getting into close combat. Kill your enemies without needing to deal with the vagueness of poisons or henchmen.
Yet inexplicably, many places and societies restrict the use and sometimes even the bearing of guns. There are whole sports built around the fiction that you can't just shoot your opponent. How does this fit together?
Clearly, this is just societies avoiding a tragedy of the commons, an inadequate equilibrium. Your life might be slightly improved by you carrying, but it will be massively impacted if everyone around you (including kids and schizophrenics) is likewise packing. The gun CW is merely about the specifics of the tradeoffs, but very few people support just arming everyone.
As I have outlined above, permitting LLM-generated texts would transform this website in pretty much the same way as airdropping a few crates of Kalashnikovs would transform schoolyard violence.
I'm not sure I agree with the point it makes, but that's a really cool analogy.
More options
Context Copy link
More options
Context Copy link
LLMs shouldn't be opining on politics in the same way paid shills shouldn't, there's something insincere about it. They're explicitly trained to have a political stance. They reflect on the companies that make them after all. They are like paid employees and not free citizens. Talking politics or arguing with a paid employee about politics is no good.
Whereas having someone program something for you is standard.
How can the achievements of LLMs in programming be denied? It's an objective fact, they can do these things.
Also I genuinely don't think that they're dumb or boring outside of programming. I could talk with an LLM for an hour about history, economics, technology, finance or fiction and it's suddenly near midnight. There is a lot of depth in long-context and long prompts where one can get out of 'surface-level' talk. Unfortunately, the main use-case is agentic coding where prompt adherence is very important. So the creativity can be hard to access as it rigorously tries to obey what you told it to do.
I don't think I would mind talking to someone being paid to advocate for a political position, as long they were honest about the situation. I wouldn't mind talking to a worker for a political party, the same way that I don't mind talking to party volunteers outside polling places on election day. Still, you may have a different attitude to me.
I'm not denying that LLMs might be useful in programming. I'm not a programmer myself - I neither know nor care about the specifics of that field. What I deny is that usefulness for coding clearly translates into domain-general usefulness. As far as I can tell bots are useful for some forms of symbol manipulation, but are quite limited and narrow in their usefulness beyond that. When they manipulate symbols in fields that I'm familiar with, i.e. words, I find they are consistently bad at it.
I did see something going around about LLM intelligence, something along the lines of, "compare Fable to your ten smartest friends", with the implication that Fable would easily win that comparison. All I can say in response to that is, "get better friends". I've asked bots about history, philosophy, theology, fiction, politics, and in general things that I consider myself to know a lot about, and they are invariably awful. There is just no point asking one to do anything.
More options
Context Copy link
More options
Context Copy link
I wrote a whole post about this.
Whether or not AI text is "worthless slop" depends on context. For fiction, or original thinking? Yes. For summarizing things or presenting information? No. The reason we prohibit LLM-posting (except when clearly stated as such and to serve a purpose) is that the Motte is supposed to be a discussion site, and no one (well, almost no one) actually thinks LLMs are "discussing" anything. If you just want to see LLMs arguing back and forth, there is OpenClaw for that.
Yes, and no. AI is already very, very good at some things. This is undeniable. It is getting better at a lot of things. It is an enormously useful tool, too useful to imagine that it's going to go away or turn out to be a passing trend.
Whether it's actually going to achieve
AGP(ETA: AGI, wtf kind of Freudian slip was that?) or trigger the Singularity is what warrants skepticism.Most AI boosters do not, in fact, think it's going to become a "super-intelligence that will obsolete us."
...I know that's a typo, but I'm sorry, I'm just tickled by the idea of an LLM successfully achieving autogynephilia.
More options
Context Copy link
More options
Context Copy link
I mean, yeah. It does. My personal experiences with LLMs (both at work doing programming and reading people's LLM-generated essays) have been profoundly unsatisfying. Therefore, I have extremely low expectations for them doing anything else well. But I think the people you see hyping up LLMs are not meeting your first criterion. For example, self_made_human has been pretty consistent in his opinions that he finds LLMs to be impressive at all the various tasks he sets to them, including writing. So it makes perfect sense that he is upbeat about their prospects in other areas. And I think the same goes for our other posters who are LLM fans - they find the models to be super impressive and therefore think they will continue to impress in the future.
I suppose my bafflement needs a top-level post at some point. It's easy to read something like this and start to feel nervous, and then I interact with an LLM for ten minutes and I am cured. These things are garbage.
Maybe it's just a matter of different fields, or different priorities. I'm looking for creativity, insight, and personality, whereas some the people worried about them are looking for... I don't know, writing computer code? Maybe in five years I'll be turned into a paperclip all while continuing to sneer that the best these systems can do is beige idiocy. LLMs don't need to be worthwhile to be a threat, and maybe garbage will destroy me one day.
Also @ChickenOverlord, please keep in mind that the cheap consumer AI you use to do your white-collar work for you or to compose an email is not the same as the one being developed by multibilliondollar companies to push the limits of technological possibility and/or conquer the world.
At the same time,
reads like a bunch of hot air about a bunch of hot air. I can't take those alignment people seriously.
More options
Context Copy link
I can echo that bafflement. Just started a new job and, like the old job, LLM usage is mandatory. And I can't just melt into the crowd like O could back at my megacorp. Every day I alternate between "this is nice and convenient" and "how did it make such a simple and obvious mistake?" It helps/hurts that most of what I'm doing is Python backend and Typescript frontend, both of which I don't care for, so I assume there's some Gell-Mann amnesia going on too. Though I'm still working on what is essentially a basic bitch CRUD app, I've yet to try it on anything truly complex.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
Because the reason the Motte is pro-tech has a lot to do with the average occupation of its users, and that occupation is what the average LLM company has been actively targeting for replacement.
It's not that much more complex than that.
It's not about to replace me for a long while yet. I am merely pivoting from code monkey to LLM wrangler. And 90% of my job as a software dev at a respectable German company is attending meetings, submitting forms for approval, and pretending to care about the minutiae of regulations. Thank God for our buerocracy; our economy might be doomed in the long run but it's good insurance against human obsolescence.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link