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 -
I think current LLMs are not remotely reliable enough to serve as politicians. Adversarial examples are a thing, after all. If we can not train an LLM to reliably avoid saying bad words, how can we expect it to reliably not vote for bad laws? And anything of truly human level intelligence would get us into alignment territory.
Then there might be a game theoretical cost if your opponents can just run your executive to determine what the reply to a provocation would be. If PresGPT was the head of the US executive, and China got their hands on a copy (one of the backups, or just a replication of the fine-tuning used on GPT5), they could have hard numerical probabilities on what the US response would be if they attacked Taiwan.
The problem with tyrants is that is attracts exactly the wrong people for the job even more than the office of president does. When Rome switched its political system from Republic to Empire, they certainly increased the variance of their leadership a lot.
And precommitting to following the policies of an assassination victim removes the incentives to kill them from the opponents of the policies, but might provide new incentives of supporters to throw them under the bus. I mean, if it is public knowledge that policy X will supported by politician P will pass with some probability p, then all you need to do is make sure that p does not change if they are killed. In practice, there is no such common knowledge, so situations where one side could act on private information will dominate.
And depending on the capabilities of the assassins, changing the mind of the successor on the policy might not be even their end goal. The end goal could be to change the mind of changing the mind of the one who succeeds the fifth murder victim or something.
Adhering to anti-racist parlance is harder than making good law, because anti-racist parlance is a logically incoherent moving target of whatever Current Thing the progressive stack is mad about today, whereas "Thou Shalt Not Murder" is pretty consistent.
Good law is a moving target, too. At least in our world of imperfect information.
And current AI does not care about logical consistency. It may usually say something consistent, because the humans who generated its training data usually tried to make their statements flow from one to the next. Unless they were being sarcastic or subversive.
Nor is there a guarantee that they were correct in their own logic. For example, they could have been starting from the assumption that things they don’t like are logically incoherent! Then, even if the AI decided to replicate their beliefs perfectly, it too would be swinging at strawmen.
That’s not a recipe for good law.
More options
Context Copy link
"Don't hand out recipes for methamphetamine" sounds pretty straightforward and coherent, though, much more than "Don't Murder", which per Wikipedia "is the unlawful killing of another human without justification or valid excuse committed with the necessary intention as defined by the law in a specific jurisdiction."
I mean, some of the best (worst) CW stories are about a killing where all sides more or less agree on the facts but their interpretation of what they think is (or should be) the law is different.
"Don't kill any humans, directly or indirectly, ever" might be simpler, but to phrase it so that our AI can't lock up people and let them die of thirst without it also being compelled to round up people and force them to take their cancer screenings or stop smoking or whatever will be complicated. There is every reason to believe that our collective ideas about these things is not particularly coherent either.
But the beauty of machine learning in general and LLMs specifically is that our ideas don’t have to be logically coherent. Which is just as well, because they never are.
You don’t have to spend ten years automatically coming up with a perfect definition of murder, you just collate a synopsis of all the people we charged for murder in the last 50 years and say, “These guys are murderers. Being like them is bad.”
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link