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 -
Neural Networks Are Already Clean
Nick Cammarata (Goodfire AI; formerly of OpenAI) claims to have discovered that earlier foundational research on the semantic structure of neural networks (especially "superposition") were wrong, and they actually (and necessarily!) work pretty much the naive way everybody used to expect them to: particular neurons mean particular, ultimately interpretable things.
If it holds up strongly, my (admittedly jumpy and currently highly-sperged) gut instinct is that this could be epochal, and a candidate for one of the more significant and re-foundational public discoveries since AIAYN if not beyond.
Yudkowsky is not impressed though:
https://x.com/alltheyud/status/2110853001749594363?s=46
Best to go one quote posts to see his replies to Anthropic researchers
Even if something does lower Yud's p(doom) you shouldn't expect him to be honest about that. I think he's being outwardly honest, though, you can see the rationalizations churning in his head - starting with "well Roon you just don't understand how my super special theory is different and unfalsifiable", giving the most uncharitable reading of the paper possible to minimize it, and then finally groping for an analogy. But Yud has a book to talk now, and human beings are very good at convincing themselves to talk their book.
I'm not really surprised by rationalists being like this. First off, it was a movement which explicitly selected for people who struggled with naturally, what's the phrase, "overcoming bias" (non-rationalists who do this naturally do it in very fuzzy ways, the classic intelligence/wisdom thing). Secondly, as TLP likes to say, knowledge is a defense against change.
Maybe I’ve just been swayed by the Rationalists’ PR campaign, but I’d wager that the average Rationalist is better at overcoming bias than the average human (which is not to say that they’re immune to it, far from it); or at the very least, people who are naturally better at overcoming bias are more likely to be drawn to Rationalism as a hobby or social group.
The average human is very bad at overcoming bias, and the Rationalists are perhaps not much better, but I’d still be inclined to put them on the upper end of the bell curve. The average person just does not think about this stuff at all, so they’re usually just going to unreflectively reify whatever biases they walk in with.
The average human is really really dumb, but there are also plenty of normal people who are not like that and didn't need to spend years reading LessWrong to get there. The positive side of the Rationalist coin is that it was a really effective program for self-improvement for nerdy tech people to deal with their struggles with overcoming bias, which helps them thinking about the things they're building, which have turned out to be the most important things in the world. But it's like saying people in Alcoholics Anonymous are really good at not drinking.
Ahistorical ragebait slander. "Overcoming Bias" was not a support group for people who were knowingly below average at overcoming bias.
"Oh they just happened to stumble into being foundationally early to the most important thing that has ever happened that even now normies are still denying, but it was just a coincidence, they're actually worse than a dentist at evaluating reality"
This comment will age poorly if we are still alive in 2030 (I am p(!doom by 2030) ~= 99.9%) given that the rationalists have claimed, much like global warmers, that we will not be, thanks to GPT 8 or whatever.
In fact, I think there will still be JavaScript jobs in 2030. Maybe fewer, but > 0.
Rationalists downloaded their AI takes not from Yud's watered down and microwave reheated logical positivism & probability theory posts, but from sci-fi. That's why every correct idea they have (AI will be important) is wrapped up with fantasy like p(doom by 2030) > 25% and « a smart AI will break out of its box by perfectly manipulating humans by discovering the magical cheat code combo of words to string out that breaks human brains to make them perfectly compliant » (which Scott Alexander recently said was a strawman of his position, but which I'm pretty sure these people, not him but people in his status hierarchy that he works with, literally used to argue in the 2010s. He hates facing consequences for his status decisions like being held accountable for views that he tolerates for such purposes).
If it were real rationality they would have a rational view of AI, which I try somewhat hard to have. They have an extremist view of AI.
I am talking about AI, not doom.
I am not sure how you don't understand the relationship between AI, and AI-doom, and how that fits into what I wrote, but whatever. Read this again maybe:
In other words I agree completely with @Bartender_Venator 's response to you, and will add you are committing a logical fallacy I have seen before, wherein you attribute what is correct but not original to a thinker you like spuriously, and ignore what is original and not correct, particularly when most of what is not original is correct and most of what is original is incorrect. That is what is going on here.
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
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link