Open Thread 434 - by Scott Alexander - Astral Codex Ten
TL;DR · AI Summary
Astral Codex Ten's Open Thread 434 features a mix of community discussions, AI-related musings, and a humorous take on tech culture, with a focus on AI ethics and community engagement.
Key Takeaways
- AI Ethics Discussion: A reductio argument about shrimp welfare and effective alt
- Community Engagement: ACX has multiple platforms for discussion and content shar
- Tech Culture Humor: A funny take on the San Francisco tech scene and AI discussi
Outline
Jump quickly between sections.
The article introduces ACX's open thread and AI-related topics.
A reductio argument about shrimp welfare and effective altruism is presented.
ACX has multiple platforms for discussion and content sharing.
The article presents a humorous take on the San Francisco tech scene and AI discussions.
Mindmap
See how the topics connect at a glance.
查看大纲文本（无障碍 / 无 JS 友好）
- Open Thread 434
- AI伦理讨论
- 还原论论证
- 社区互动平台
- ACX平台
- 科技文化幽默
- 旧金山科技圈
Highlights
Key sentences worth saving and sharing.
AI Ethics Discussion: A reductio argument about shrimp welfare and effective altruism is presented.
ACX has multiple platforms for discussion and content sharing, including Reddit, Discord, and bulletin board.
The article presents a humorous take on the San Francisco tech scene and AI discussions, including Dasani water and AI technology metaphors.
Open Thread 434 - by Scott Alexander - Astral Codex Ten
Open Thread 434
...
May 18, 2026
This is the weekly visible open thread. Post about anything you want, ask random questions, whatever. ACX has an unofficial subreddit, Discord, and bulletin board, and in-person meetups around the world. Most content is free, some is subscriber only; you can subscribe [here](https://astralcodexten.substack.com/subscribe?). Also:
- * *
1:Primary elections are coming up in many US locales. If your ACX meetup group wants to create a voter guide (samples from last election here), I’m happy to signal boost it here if your city is big enough and has enough likely ACX readers. Email me if interested.
2: New subscribers-only post, Every Magazine Piece On The SF AI Scene:
On a chilly San Francisco day, I found myself in a “group house”, a house where multiple people live in the same house. My host, a bespectacled man named Theodore Wong, grabbed a Dasani water bottle and half-eaten burrito from the refrigerator (a device used to keep food cold - according to Wong’s health-conscious bio-hacker roommate, it “prevents spoiling”). Then he started to speak.
“There is a serious risk that artificial intelligence is going to kill everyone in the world. You really need to explain this to your readers. We are desperate to get this information out. I’m begging you, please don’t make this a human interest story like all the others. We need for people to know the exact arguments,” he intoned nerdily, and proceeded to give me a twenty minute lecture including lots of words like “compute”, “exponential”, and “slaughterbot”.
This was my introduction to the wacky world of the San Francisco tech scene, where people drink Dasani water and say things about AI - a technology which, they tell me, “no really your readers need to know this, stop looking at my water bottle and listen to what I’m saying, I beg you.”
3:An Inkhaven-like writing/blogging residency in Barcelona, late Aug - early Sep.
- * *
#### Subscribe to Astral Codex Ten
By Scott Alexander
P(A|B) = [P(A)*P(B|A)]/P(B), all the rest is commentary.
By subscribing, you agree Substack's Terms of Use, and acknowledge its Information Collection Notice and Privacy Policy.
[](https://substack.com/note/p-198257206/restacks?utm_source=substack&utm_content=facepile-restacks)
Previous
#### 60 Comments
[3m](https://www.astralcodexten.com/p/open-thread-434/comment/261101605 "May 18, 2026, 4:46 PM")
I had an interesting discussion about reductios. I feel in general, without good quality domain level knowledge, a reductio is probably false, on average. There's nothing wrong with looking at the gestalt of someone else's worldview, smelling a rat and communicating that unease to that person and to others. However, it is unlikely you will correctly identify the feature of the system that is causing the unease. Any moral system is somewhat complex and "the tail may come to wag the dog".
Example: I think shrimp welfare is weird, I'm inclined to smell a rat, but if I were to make a reductio - "if you were effective altruists you would care about people, not shrimp", that assumes I have correctly identified the final cause of effective altruism - human welfare - and can therefore judge the efficient cause - shrimp charities - as contradicting the final cause. If I know my stuff about EA, I stand a chance of doing this in a way that might convince some fence sitters, but in lieu of this I'm essentially firing random shots. My opponent felt I was being too humble and that cliques are prone to over-fitting and "circle jerking" Any thoughts welcome.
[16m](https://www.astralcodexten.com/p/open-thread-434/comment/261094233 "May 18, 2026, 4:33 PM")
If you are excited about biosecurity or biodefense you should consider applying to the Frontier Biodefense Fellowship. The program is one of the very few opportunities to get started working in biodefense research, policy, entrepreneurship, especially if you haven't had tons of relevant experience yourself!
Mentors are from virtually all orgs that care about scope sensitive biodefense: SecureBio, Blueprint Biosecurity, CLTR, Coefficient Giving, and many more!
Deadline is June 2nd, apply at fellowship.bio!
27m May 18, 2026, 4:22 PM
Does anyone have a good explanation of the US stock market's performance during the Iran War? I'm torn between two different observations.
#1 John Mearsheimer (1) and others predicted massive oil shocks as a result of the Iran War which would lead to recession. This seems largely to have occurred. Oil prices have risen significantly all over the US and everyone I talk to is feeling it at the pump. Seems like it should be a massive oil shock.
#2 Stonks go up. The S&P was $6,858 at the start of the year, ~$6,878 at the start of the Iran war, bottomed out on 3/30 at $6,343, and is currently at $7,383.
I notice that I'm confused. There's a massive risk to global oil markets, that risk materializes, over the next month, the market declines, and then at the end of March the market bounces back $1,000 over 6–7 weeks despite the oil crisis getting worse. On the one hand I'm seeing pretty credible evidence of an oil/energy shock and yet stonks go up. I do not understand this.
(1) As an example of an excellent IR resource, not some rando streamer.
Oil shocks don't have a major impact on the earnings power of companies whose good fortunes have been the engine of the S&P 500 since 2023.
... are you perchance under the delusion that there isn't active governmental interference trying to keep the global financial markets running? Wheels are close to falling off, so it's pretty stressful down in the trenches.
America is making out against China, in the energy "war", so of course our stocks go up (stopping the gulf oil really kicks China in the bullocks)
The real crisis is being deliberately kept under wraps, so it's no surprise that you're looking at oil instead of water.
Yeah, new email addresses are free.
Edited
Grimmoar/gawdflea/3 other banned nicks, could you just stay away? The conspiracy theories and hints at hidden information are really annoying enough, but combined with the flood-style of posting and derailing it's dragging the comments down. Please, there is a reason for the previous bans.
Wimbli—>Aristocat—>Zanni—>Gawdflea—>GrimMoar
Let’s play Where’s Wimbli.
That post contains a recommendation for Emily Oster's "Expecting Better," as a good evidence-based guide to which parts of general pregnancy wisdom are real and which are BS; it looks like there's a 2025 update to that: https://www.amazon.com/Expecting-Better-Conventional-Pregnancy-Wrong/dp/0143125702
[31m](https://www.astralcodexten.com/p/open-thread-434/comment/261086164 "May 18, 2026, 4:19 PM")
I'm not a biologist, nor a zoologist, but I watch pop science videos about extinct critters. And there's one thing that bugs me immensely about phylogeny.
Everyone seems deathly afraid of actually labeling any group of species X as the ancestor of another group of species Y. Don't believe me? Search for "phylogenetic tree of life", and you'll mostly find depictions like this one: https://www.researchgate.net/publication/252743392/figure/fig5/AS:365282104233990@1464101448858/The-phylogenetic-tree-of-life.png
Apparently, if you lived more than 1 million years ago and among your countless descendants are representatives of two different species, then you don't belong to any class, order, or family. Your genus shall not be named, and your place on the tree of life shall forever remain an unmarked fork on the evolutionary path. Fish aren't shown to be descended from archaea, amphibians from fish, mammals from amphibians. The tree always branches off just before.
Except for birds and dinosaurs. Because birds totally are dinosaurs! Oh, sorry, I meant to say "avian dinosaurs". Yeah, that's how totally birds are dinosaurs. And therefore a sparrow and a triceratops are both dinosaurs, despite branching off from each other some 230 million years ago. Fuck this!
[9m](https://www.astralcodexten.com/p/open-thread-434/comment/261098352 "May 18, 2026, 4:41 PM")
The whole point about evolution is that there's a smooth transition from one lifeform to another. You can dice up the gradient of life into "phyla" and "species", but your categories are going to be fairly arbitrary no matter what you do. Outside of massive divisions such as "this thing can absorb sunlight for energy, and this other one can't", the fine-grained details will always come down to historical precedent, a.k.a. vibes.
[23m](https://www.astralcodexten.com/p/open-thread-434/comment/261090614 "May 18, 2026, 4:27 PM") Edited
IIRC the classic taxonomic system of orders and families has been deprecated for about a decade. It's less that they don't bother naming new families anymore and more that there shouldn't have been any families named in the first place. But I'm no biologist, don't quote me.
The birds are dinosaurs thing is less about saying how cool birds are, and more about highlighting the fact that actual dinosaurs of old were in fact feathered. I also think it predates the deprecation of taxonomic ranks by like 6 months.
[38m](https://www.astralcodexten.com/p/open-thread-434/comment/261081993 "May 18, 2026, 4:11 PM")
I've stopped watching war coverage. But I check the oil price, and that gives me a hint. It's getting worse in the 'oil prices' opinion.
[41m](https://www.astralcodexten.com/p/open-thread-434/comment/261080696 "May 18, 2026, 4:09 PM")
I agree, but I think there's a difference between a life story truly necessary to produce the work, and a life story that merely makes the work seem more interesting once you know it. If we put too much weight on the latter, then art becomes basically a genre of celebrity trivia and memorabilia. "This is the actual katana that Mishima stabbed himself with after he tried to overthrow the government! Oh and this is the actual novel that Mishima wrote before he stabbed himself with said katana! And this is a tissue he once sneezed on!" At a certain point you have to let the novels stand on their own. In Mishima's case, I think they do for the most part.
[35m](https://www.astralcodexten.com/p/open-thread-434/comment/261083844 "May 18, 2026, 4:15 PM")
I don't think those are the same at all. An artist's lived history MAY influence the piece (Snow Crash is very influenced by the Author's rebellious nature, but you have to read Diamond Age to understand that).
... or are we not allowed to talk fictional authors? (aka fanfiction?)
Art is done in relationship to other art, and this is best seen in parody, which cannot be understood without reference. Not all art stands alone (in fact, little art does stand precisely alone -- Snow Crash could be seen as poking fun at Neuromancer and Gibson). That's different though, than the artist's own history (beware: you may not read someone's real history, depending on the author, in which case, do you find the novel still as good?).
I would like to say that if you can only understand a novel because of "where the person is standing" as they're writing it, that's probably a pretty bad novel (I hold some level of universal standard, like how I expect Big Bottom or other parody songs to Actually Have Good Music).
[27m](https://www.astralcodexten.com/p/open-thread-434/comment/261088167 "May 18, 2026, 4:23 PM")
I agree they are different in the ways you describe, but I'm trying to get at the fact that knowledge outside the piece of art itself can change the way it's viewed. This could be knowledge of personal history or art history. Good art should be digestible without knowing the personal history, but it can change the experience.
I feel like the piece on taste was dismissing this. That if you look at two pieces of art in a vacuum and like one better, it seemed to be making the argument that knowledge about the history or creation of the art shouldn't influence your opinion. You should just keep liking the same one, because your "taste" implied it to be the better painting/book/poem/etc. E.g. the part about Chestertons poems.
But the history does matter. It changes my emotional state when absorbing the piece, which is the basis of my taste.
[18m](https://www.astralcodexten.com/p/open-thread-434/comment/261093270 "May 18, 2026, 4:31 PM")
I still feel that you do not distinguish "history of the author" and "history in general" (or, perhaps better put, simply "context") enough.
[1h](https://www.astralcodexten.com/p/open-thread-434/comment/261078249 "May 18, 2026, 4:05 PM")
Believing there's a God who loves you solves a concurrency problem.
Just wrote this, would love thoughts or feedback. This is especially relevant to people interested in predictive processing models.
https://apxhard.substack.com/p/monotheism-solves-a-concurrency-problem
So he wants to break the rules, but he wants permission first. There's this strange desire among these people to be both one of the good kids and one of the bad kids at the same time. The fact that he's 40 makes it stranger.
If I were in charge I probably would demolish the ballroom, but it would be relatively low priority, and I'd go through all the proper approvals and permits and such.
[34m](https://www.astralcodexten.com/p/open-thread-434/comment/261084287 "May 18, 2026, 4:16 PM")
This is the guy who got kicked out of antifa.
And, no, you wouldn't demolish the ballroom, if you were in charge. If you were in charge, you'd know why it got built.
[32m](https://www.astralcodexten.com/p/open-thread-434/comment/261085363 "May 18, 2026, 4:18 PM")
I don't think he was ever really part of antifa. He's a Twitter-turned-Bluesky progressive.
[42m](https://www.astralcodexten.com/p/open-thread-434/comment/261079975 "May 18, 2026, 4:08 PM")
I know you're joking, but terrorists tend to come from the anarchistic direct-action "propaganda of the deed" strain of leftism rather than the bureaucratic vetocracy strain which takes hold in established governments.
[42m](https://www.astralcodexten.com/p/open-thread-434/comment/261079775 "May 18, 2026, 4:07 PM") Edited
This is not how terrorist cells operate.
[1h](https://www.astralcodexten.com/p/open-thread-434/comment/261064512 "May 18, 2026, 3:40 PM")
I thought of a strange conciousness/pzombie thought experiment involving two previous ones. The first is blockhead, a computer that seems to be concious because it has one GIANT lookup table for IO pairs of human conversation. Note that Blockhead inputs could technically even include vision, audio, etc. while outputs could include motor responses, etc, and be arbitrarily quantized regarding time (1 second of IO, or 1 millisecond, etc) It is often used as a modus tolens against behavioralism, because obviously a lookup table isn't concious! Chalmers Dancing Qualia is another thought experiment where you go through the whole brain, taking neurons and converting them to a silicone equivalent, preserving functionality. As you go, the subject should report if qualia disappeared, but if they did, that would be a loss in functionality! Many accept this as a good argument against substrate dependence.
Well lets take a human brain and a single neuron in it: replace it with a lookup table for inputs and outputs. Now replace an adjacent neuron with another table. Now combine those two tables into one larger table. Keep doing this in any order you see fit, until you have one giant lookup table. We have danced our way from a human brain to Blockhead. Remember, any change in concious experience during the process should have been reportable. Any change in reported qualia would violate the functional requirement. Where on the subdivision dimension did it start/stop becoming conscious? Some people try to get around the Blockhead question by claiming that only the thing that recorded it's IO pairs must have been concious. This dancing process closes that loophole somewhat. My conclusion is that a true Blockhead (a lookup table far, FAR larger than our universe btw) would be as concious as I am, OR there is problem, other than size, preventing a lookup table from ever functionally cloning my brain, like how a non turing complete language would be unable to emulate a turing complete one.
[13m](https://www.astralcodexten.com/p/open-thread-434/comment/261096056 "May 18, 2026, 4:36 PM") Edited
The problem with all those Chinese Room Blockhead p-Zombie thought experiments is that they're totally unfalsifiable. If you ask the Chinese Room/etc. if it's conscious, it will say "yes", because that's what its loopkup table indicates; and in all other respects it acts as though it were conscious, so there's no way for you to tell that it isn't really truly conscious unless you take it apart -- and maybe not even then. This leads to the following possible conclusions:
1). Only biological systems can be conscious, because I have a priori decided this to be axiomatically true.
2). Consciousness is mysterious and ineffable and nonphysical and it's all a grand spiritual mystery.
3). The philosophical concept of "consciousness"/"qualia"/etc. is hopelessly confused, on par with "elan vital" but more so.
Personally, I'm going with option (3), but that's just me.
[1h](https://www.astralcodexten.com/p/open-thread-434/comment/261072870 "May 18, 2026, 3:55 PM")
Many accept this as a good argument against substrate dependence.
I don’t know why people don’t think about the mind being analogous to software here. Something running on a system (or substrate) that depends on the lower system to run but isn’t itself the lower system. In the case of a future AI living in a data centre we could replace all the data centre machines, and keep the AI. If it felt qualia to begin with it it would feel it to end with.
[1h](https://www.astralcodexten.com/p/open-thread-434/comment/261069515 "May 18, 2026, 3:49 PM")
a lookup table isnt concious
This is basically the Chinese Room thing again, and I find it kind of annoying (no offense to you in particular). You take some data structure and simultaneously say it "obviously" is or isn't [a thing we can barely define] but then you also extend it to a level of complexity we can't really understand (arbitrary size, all kinds of inputs, the passage of time) so it doesn't even really resemble what we would normally think of as "a lookup table".
With these constraints you could implement a deep learning model as "a lookup table" and therefore approximate any continuous function.
[14m](https://www.astralcodexten.com/p/open-thread-434/comment/261095839 "May 18, 2026, 4:36 PM")
BTW is a lookup table the same as sin(x)?
(Not sure about the answer, just asking your opinion.)
[5m](https://www.astralcodexten.com/p/open-thread-434/comment/261100530 "May 18, 2026, 4:45 PM")
You could implement it that way. I'm pretty sure I had at least one textbook when I was younger with a few pages of sin(x) in the back.
You'd have some limitations - for the paper version, only accept inputs to some finite number of digits and bound it to some 2pi range.
If you rule out those limitations, you'll need a really big table, but it's unclear what the limit would be in the thought experiment given.
Have you seen this yet, and will you add your name to it? Your article about why UBI is superior to a jobs guarantee remains evergreen and great to this day. https://actionnetwork.org/petitions/the-ai-pledge-for-humanity
[1h](https://www.astralcodexten.com/p/open-thread-434/comment/261056119 "May 18, 2026, 3:25 PM")
In the current debate with the buttons: https://knowyourmeme.com/memes/red-button-or-blue-button-question
At what percentage of inherent disadvantage to your default answer do you switch sides?
I'd press red in the original setting, but if a blue outcome was allowed with just 5% of blue votes, I'd easily choose blue. I can't know for certain but I think my boundary is around 30%
[1h](https://www.astralcodexten.com/p/open-thread-434/comment/261059821 "May 18, 2026, 3:32 PM")
The problem is so reflexive that even just the manner in which it would be presented has a strong influence on what everyone might think about everyone else's thought process, and thus how one 'ought' to answer. The attempts to transform the question by imagining jumping into a woodchipper which disables if 50%+ of people jump in because the question is not merely whether you, the reader, are willing to jump in or push red, but what you think the likelihood is that others will do so. In the abstract I'm a red-pusher with the original framing but if the presentation of the threat to red-pushers were sufficiently menacing it would shape my decision.
[1h](https://www.astralcodexten.com/p/open-thread-434/comment/261069235 "May 18, 2026, 3:49 PM")
Oh, yeah, I've been having this issue and it's pretty infuriating. I also hate the stupid tribalism it's spawned, and this question comes from a need to mend that tribalism. For the moment let's say it's the exact same framing with just the percentage required for a blue win adjusted.
In the lower percentages (say, blue wins with 5%), i'd guess the threat to pushing red is social stigma or shame, in proportion with how low the threshold was.
[1h](https://www.astralcodexten.com/p/open-thread-434/comment/261077677 "May 18, 2026, 4:04 PM")
The button isn't real. But the hate, anger and contempt that people feel for each other, over a thought experiment, is real.
That right there seems to be a way that evil can grow in the world in ways that most of us miss.
What time/time zone on May 20th are the Book Review submissions due? Asking for a friend 😅
[2h](https://www.astralcodexten.com/p/open-thread-434/comment/261051176 "May 18, 2026, 3:16 PM")
Anyone with connections to Hackensack Meridian Health?
[2h](https://www.astralcodexten.com/p/open-thread-434/comment/261049825 "May 18, 2026, 3:14 PM")
The Four Kinds of People in the AI Era:
In the Age of AI, four quadrants emerge that tell four very different stories about where people are headed. There's the Augmented Thinker, the Cognitive Outsourcer, The Traditional Thinker, and the Passive Drifter. Which one will you be?
https://markmcneilly.substack.com/p/the-four-kinds-of-people-in-the-ai
[9m](https://www.astralcodexten.com/p/open-thread-434/comment/261098554 "May 18, 2026, 4:41 PM")Edited
Okay, I'm probably Quadrant 3 but ignore that, I want to fight with you over this:
"Pericles, the great Athenian statesman, said, “You may not be interested in politics, but politics is interested in you.”
I went "I never read a word by Pericles but I'm damn sure he never said that" and going back to the original source, insofar as I'm directed to an original source, no, indeed he did not:
https://home.uncg.edu/~danford/pericles.html
"It is wrong to say that the people love Athens because she is great. It is the other way around --- Athens is great because the people love her and take part in politics. Here each individual is interested not only in his own affairs but in the affairs of the State as well: even those who are mostly occupied with their own business are extremely well-informed on general politics; this is a peculiarity of ours; we do not say that a man who takes no interest in politics is a man who minds his own business; we say that he has no business here at all.
Pericles
Speech to the Athenians in 450 B.C."
So "Politics is interested in you" is a stupid, modern, goldfish attention span, Internet quotes (worst of) boiled-down version of what he did say. It's not "You may not be interested in politics but politics is interested in you", it's "Athenians believe it to be their civic duty to take an interest in the affairs of the State, and this is what makes Athens great".
You have just demonstrated why I'm never going to use AI to think for me or help me think if I can possibly help it; we already have bad Internet Quotes where historical figures are represented as saying shit they never said (pardon the strong language, I feel deeply about this) and AI-mangled précis of data scraped off the Internet is only going to make it worse.
Go back to the original sources or nothing!
[1h](https://www.astralcodexten.com/p/open-thread-434/comment/261073910 "May 18, 2026, 3:57 PM")
I’ll be in the local pub on a Thursday doing the pub quiz. Graham Norton turned up once.
[2h](https://www.astralcodexten.com/p/open-thread-434/comment/261050853 "May 18, 2026, 3:16 PM")
What reason is there to think those labels are carving nature at the joints rather than being MBTI type fabrications?
[5m](https://www.astralcodexten.com/p/open-thread-434/comment/261100873 "May 18, 2026, 4:45 PM")
It's a guy saying "MY job is safe, I'm an Augmented Thinker!" while using AI to help him write the piece, seeming not to recognise that AI might produce such in future without the need of a human middleman to generate prompts.
The irony may turn out that the job of the fella sweeping the floors in the university is safer in the AI future than that of "Mark McNeilly is a Professor of the Practice at UNC Kenan-Flagler Business School and chairs the UNC Provost’s AI Committee. He writes on AI, leadership, and strategy at Mimir’s Well on Substack.
This article was written by me and improved with additional input from AI."
[2h](https://www.astralcodexten.com/p/open-thread-434/comment/261050439 "May 18, 2026, 3:15 PM")
I was planning on #5: The Paperclip
[2h](https://www.astralcodexten.com/p/open-thread-434/comment/261051780 "May 18, 2026, 3:17 PM")
Satisfy values through friendship and ponies!
[2h](https://www.astralcodexten.com/p/open-thread-434/comment/261049090 "May 18, 2026, 3:13 PM")
Did anyone find the voter guides useful? (as in they weren't 99% what you would have guessed without seeing the guide?)
[2h](https://www.astralcodexten.com/p/open-thread-434/comment/261051876 "May 18, 2026, 3:18 PM")
It's often helpful when they ask candidates questions about some topic, or organize candidates' public stances on those topics.
[2h](https://www.astralcodexten.com/p/open-thread-434/comment/261048169 "May 18, 2026, 3:11 PM")
Is the excerpt from the subscriber article a quote from a real article that is being critiqued, or a parody? I honestly can't tell
[1h](https://www.astralcodexten.com/p/open-thread-434/comment/261074391 "May 18, 2026, 3:58 PM")
It’s a parody written by Scott.
[2h](https://www.astralcodexten.com/p/open-thread-434/comment/261052621 "May 18, 2026, 3:19 PM")
Poe's law is brutal...
Top Latest Discussions
Sixty-eight years of highly defective people
Jan 16
2,063
You just keep on trying till you run out of cake
Jan 21, 2021
1,589
