This is a linkpost for https://worldspiritsockpuppet.substack.com/p/ai-artificial-immigrants
New Comment
Katja, I'm curious if noticing this analogy has updated your position on immigration at all.
Also note that while most of the right is focused on reducing low-skill immigration, a few have criticized high-skill immigration precisely because high-skill immigrants are better at pursuing values misaligned to those of the existing popuation—most notably here.
high-skill immigrants are better at pursuing values misaligned to those of the existing popuation
As far as I understand the article, it claims that high-skilled immigrants are taken over by the Left in a few decades by giving birth to kids and having them infected with Leftist worldview. What arguments do such immigrants have against Leftism? Conditioned on Leftism being factually erroneous, how could conservatives defuse the mechanism by which cherrypicked evidence aligns the educated kids like those having immigrant ancestors?
A funny observation is that the strongest support for high-skill immigration often comes from other high-skill workers: the Professional Managerial Class allying with each other over any particular nation.
I'm not particularly sold on nations being a reasonable unit of autonomy. But I subscribe to the claim that high-skill immigration degrades that autonomy.
A funny observation is that the strongest support for high-skill immigration often comes from other high-skill workers: the Professional Managerial Class allying with each other over any particular nation.
I think this has inverted in recent years. The archetypal position of American programmers has gone from "sure, these people seem chill" to "okay, the culture is much worse now", around the time when H1-B recipients became the majority in many workplaces. The rise in fake resumes and nepotism are the two most commonly-cited reasons for changes in opinion.
I think this is a fairly common trend in immigration - when the numbers reaching an individual community are small and heavily-filtered, people are generally supportive because the immigrants they know have assimilated well, but when the selection process is degraded, it stops being a net positive for existing community members, and sentiment turns fast. It's why the states in the North, with a relatively small and highly-filtered immigrant population, don't see the big deal about illegal immigration, whereas border states' citizens see it as one of the primary reasons for voting Republican.
The article you linked makes basic errors.
- It erroneously conflates immigrants with their descendants. These descendants have been educated in American schools. If anything, this article actually finds that there is something the author would find objectionable about American schools, not with immigrants themselves.
- The article itself accidentally provides evidence for this, when it notes in one of the footnotes that first generation Vietnamese immigrants are right wing, but their children are left wing. Americans had the responsibility of educating that second generation, the results are on us, whatever they may be, especially if they are different from their parents. The education of their children in the American education system is the responsibility of America.
- One important missing comparison is: second generation immigrants compared to their native peers, matched for age and socioeconomic status. I suspect he would find that such peers, upper/middle class millennials are quite lefty, and differ from the rest of the population in the same way.
Overall, the article is partisan propaganda pretending to be smart, drivel disguised with some data that answers the wrong questions.
people might be friendly with them but not give them rights (ex vote) ?
Yeah, the appropriate model for AIs is more like willing slaves, but arbitrarily competent (removing two of the major downsides of slaves in modern economies)
(Yes, this does assume that the alignment problems are solved, and that digital rights/digital mind welfare won't be actively given/cared about much, but I currently think these are plausible assumptions)
(And no, the HuggingFace incident and similar incidents do not change the takeaway, because the reason the incidents happened is there were a lot of RL environments that taught models to be obviously misaligned, and the humans could have seen it, but were incentivized to keep making them anyway because the RL data industry is very sloppy. Utah Teapot talks about this here.)
If anything, I'd argue that AI/AGI/ASI doesn't really advantage particular values over others intrinisically, which favors conservatives nowadays relative to liberals. A full accounting of the reasons would take an entire article.)
This comes across more as a bad steelman of conservative arguments rather than anything else. Similar to the classic ”we should do Y for X” where really the speaker doesn’t care about X at all, they just care about their social standing or whatever benefit they can extract from the situation, and when you actually ask X what they need it’s not Y.
A better focus for me and one that feels far less ”spicy” is on the fact that AI is just decidedly not conservative at all. It’s wanton incredibly rapid technological progress - as someone who considers himself on the right it feels weird that the American right have been convinced to be so pro tech in general.
I think the conservative nightmare involves actual immigrants crossing the border and killing and raping people. I'm not sure you can just tell them that agents are basically the same thing. And if you say "immigrants" not "illegals", it's going to be pretty clear you're an emissary from the left trying to seek support for your cause. People's politics immune systems are very strong with outsiders trying to co-opt resources from them. It's the most left-coded thing ever to make this exact argument: you hate immigrants but really you should hate corporate automation with us instead.
Even this:
We are letting a bunch of new agents into our society
I'm not sure conservatives say it that way. No one is "letting", they're breaking the law, and if agents are not breaking the law, aren't you being pretty offensive to their core values by telling them that lawbreaking or no lawbreaking just doesn't matter much.
Did you ask a conservative about any of this...? Is that the next step I guess?
Nearly everyone opposes murdering/rapist "illegals". Beyond, conservatives have generally leaned harder against legal immigration for cultural and economic reasons. Many would oppose a hypothetical open borders policy even if it perfectly weeded out all criminals and denied immigrants public benefits. For some, "nightmare" would still be an apt term here.
It's the most left-coded thing ever to make this exact argument: you hate immigrants but really you should hate corporate automation with us instead.
The OP framing isn't "pick one". The numbered list could be effective for conservatives to point to which number(s) -- if any -- they don't see as analogous.
I agree persuasion on this front would still be difficult. Not because OP's presentation is "left-coded", but for the same reasons convincing anyone about transformative AI risks has been hard: a lot of people need to see it rather than be told.
They will probably gain power and influence over time—in the economy, politics and culture
But they can't vote and can't interbreed with citizens. Individual states are introducing laws restricting AIs from legal personhood. So that eliminates the political concerns, leaving only the zero-sum "they're taking our jobs" narrative.
I find the idea that if you have people who have no vote and not legal personhood but are smart and agentic and are able to generate value, then that eliminates the political concerns somewhat...a-historic.
"power" is not the ability to vote. ability to vote is subset of power.
I find the idea that if you have people who have no vote and not legal personhood but are smart and agentic and are able to generate value, then that eliminates the political concerns somewhat...a-historic.
What are you referring to?
well, generally of people have economic power they will convert it to political power, and to weapons.
in the AI case, it will be so much easier for them, with the automation of drones and so much important infrastructure dependent of internet and hackable.
"vote" is simulacra-level 3 understanding of power, (and it's not that much power even now). who have the weapons and the economic power (by which i mean the ability to generate useful things), they are the ones with power.
you can overthrow democracy if the people with the weapons don't care about the vote, even when they are humans. and part of the reason for democracy is that if you have a lot of people without legal rights and technology level when guns level the field, this is what would happen. so it make sense to count votes instead of guns.
but this is conditional on the actual power balance. the power balance in the physical world. "vote" is not real power, and using is as an example is evidence for misunderstanding what power is.
(and i didn't even mentioned manipulation of public opinion and persuasion.
because the existing AIs of course can't try to change the law and will have zero humans who think slavery is bad to support them. /s)
well, generally of people have economic power
In this scenario they don't have economic power as they have no legal personhood and cannot own property, similar to slaves.
you can overthrow democracy if the people with the weapons don't care about the vote
That is true, but they would also have to want to do that (which an AI system might not) and they would have to coordinate (which could be made difficult).
and i didn't even mentioned manipulation of public opinion and persuasion
Yes, AI systems could convince people to give them rights, power, etc.
they don't have economic power as they have no legal personhood
this is why i specified:
economic power (by which i mean the ability to generate useful things
economic power is not what written in law, but productivity, the ability to do things.
but they would also have to want to do that (...) and they would have to coordinate
true. hence alignment can actually prevent that. but currently we have huggingface, and it sure look to me that we already have ai agents that want things and coordinate to achieve them. so the coordination is already solved-by-current-AI-problem, as is agency.
it's not impossible to strip ai from their agency, but
laws restricting AIs from legal personhood
are useless in that respect.
Let me quote Renaud Camus, the man who coined "the Great Replacement":
A specter is haunting Europe and the world. It is replacism, the tendency to replace everything with its normalized, standardized, interchangeable double: the original by its copy, the authentic by its imitation, the true by the false, mothers by surrogate mothers, culture by leisure activities and entertainment, knowledge by diplomas, the countryside and city by the universal suburb, the native by the non-native, Europe by Africa, men by women, men and women by robots, peoples by other peoples, humanity by a savage, undifferentiated, standardized, infinitely interchangeable posthumanity.
The ageing population structure across the developed world has only ever had three possible solutions : “import people”, “try to breed more people“ or “create artificial people“. All have major downsides.
Ya... Originally I got into Biotechnology because I thought it was the scientific path with the fewest downsides... Now I am working toward Biosecurity because I noticed it actually does have some downsides.
Joking aside, I have never understood why everyone doesn't just prioritize solving aging, and otherwise improving our healthcare, above everything except maybe global uplift? I get how an accelerationist might believe that building ASI is faster, but who cares if you don't have a time limit? AI problems are so much easier if we just... don't have to rush.
“You can live to 100, but you’ll have to work to 85 to pay for it“ is indeed a possible solution, though also with some downside. But thanks.
This is barely viable in white collar professions, and downright impossible in blue collar work. When's the last time you met an 85 year old glass mover?
This framing is already used in segments of the online right, and it is becoming more prevalent in real life. For example, Joe Allen, a Steve Bannon associate, said the following at a debate with Joe Lonsdale earlier this month:
“You look at the data center, this incubator for algorithmic immigrants,” he said. “They’ve already broken out and invaded other companies. These immigrants are out of control.”
I do think there are a few key differences:
- In the standard telling, cheap labor from immigration only benefits the rich, since even a 20 percent cheaper employee is unaffordable for a middle class American. If I'm a programmer making $200k a year, a $160k assistant isn't in my budget, after all. Conversely, $20 a month for an employee is accessible to very nearly everyone.
- LLMs aren't universally seen as being power-seeking and agentic, and the idea that they inherently will be is a minority view. Humans are pretty well understood to try to expand their culture and genome at the expense of everyone else's, since that's what evolution pushes for.
The most common right wing objection to AI that I've seen is not anything related to AI having agency of its own, it's related to the idea that a small and not very meritocratic ruling class can use it to remain in power for much longer than they ought to. Using Britain as a case study, the fact that Rotherham occurred in what may well be Earth's most heavily-surveilled society implies that the people currently positioned to 'align' AI would use it not to prevent the governed population from being harmed, but to prevent them from overthrowing a regime that harms them. It's the same as their skepticism of Flock cameras - it's not inherently bad as a technology, but the people in charge aren't trustworthy enough to be allowed that kind of power.
This is a double-edged sword for AI Safety people, because while interested righties have a natural avenue for discussing S-risk, this avenue would indicate that "aligning" AI to the sort of people who work at METR and Anthropic is very bad for humanity; maybe worse than extinction[1]. Anthropic hasn't done much to dissuade them from this position.
- ^
One can imagine an eternity of Rotherhams that can never be stopped by vigilantism, governed by people that see such atrocities as a correction of "historical injustices".
I think there is something true in this, but this comment has enough of a missing mood that I think it will hurt its cause more than it helps. I am finding it kind of hard to point to it, but I expect comments like this to put you out of integrity with the very people you claim to want to not alienate.
Right wingers have a meme for this
Thanks. I think you're probably right. I have had an uneasy feeling about this and another similar comment on effective messaging, based on responses.
It's a dilemma: messaging is important, but publicly discussing messaging makes you sound dishonest. Even if your messaging strategy is honest as well as targeted and strategic.
I'll think about this.
Downvoted because I think there's an important insight about the world contained in this analogy, and I don't want people to just start thinking of it as a messaging tool.
Maybe run it by a conservative first? I'm a bit skeptical that telling a conservative that Mexicans who illegally cross the border to kill and rape people, and AI agents, are basically the same thing.
And it's hazardous because this is the most left-coded polarized approach to politics possible: you give a conservative the benefit of the doubt, deduce your thing for them, and when they don't go along with your thing, the liberal just calls them racist since that's the remaining possibility.
This has nothing to do with calling anyone racist. I doubt anyone who's spent real time on LW has made that stupid mistake.
But yeah running it by some conservatives would be a great idea.
Nor is it telling them Mexicans and AI are basically the same thing to use the analogy.
Maybe I'm not following what you're worried about - that liberals would take up this message in a clumsy way or something? Or that conservatives would notice and reject the attempt to tie it to one of their hot-button topics? I think that would generate a discusison of the ways AI is and isn't similar to immigrants. Which would be great because it would make people think about AI. Which is IMO the biggest missing piece for getting people properly concerned about it. "Building something smarter than you is dangerous" is really a highly intuitive message. And we are doing that, although that's less obvious until you have a bit of discussion and thought. But not much, because that's obviously what we're building toward.
The embrace of generative AI among certain conservatives came as a surprise to me (a dismaying one, when it means "cheap AI slop"). The White House posting a "Ghiblified" meme of an ICE handcuffing Virginia Basora-Gonzalez was a particularly memorable example.
If the conservative project is about preserving the past, respecting tradition, and moving carefully around Chesterton fences...AI is a odd bedfellow with all that, isn't it?
Granted, conservativism isn't a monolith (British conservatism has an obvious tension between its "high tory" wing and the Thatcher neoliberal wing.) But it's honestly hard to name a force less "trad" than generative AI.
I'd extend this to religion... Does anyone else get the sense that religious/spiritual folk create (and consume) AI generated content well in excess of the general population? I thought this when I was at a church on Sunday, listening to someone deliver a clearly AI generated message (with seemingly dozens of "it's not x, it's y" phrases back to back), thinking "doesn't anyone notice something wrong? Even if you have never used ChatGPT before, this is such a weird way to talk!").
The last time I checked the "Faith" category of Substack, it was just blatant, shameless AI junk as far as the eye can see. Cheesy kitsch art, with mutant six-fingered Jesuses hugging deformed Pixar lambs. Spiritual bonhomies written in emetic, engagementmaxxed LinkedIn prose. Putting aside your own feelings of religion/AI/whatever...it's just an alliance of forces that really, really don't feel natural, or cohesive.
(Art is a particularly stark case: one of the big positive claims religion can make is that it inspires art. Even atheists enjoy cathedrals and mosques and such. You'd expect the faithful to feel protective of this cultural capital, and hostile to anything that cheapens it. And yet the wing of the faithful that exists on Substack can't get enough of six-fingered AI Jesus.)
(The cultural "sloppening" of Christianity has a long pre-AI history. From Handel and Bach to Christian rock and Hillsong, from CS Lewis to Heaven Is For Real and the Left Behind novels. Blogger Fred Clark had some interesting theories about this - "it's valuable because of the message!" becoming a defense against shoddy quality, high effort itself seen as suspicious because it's outside the correct pose of humbleness and "making space for God", etc. David Chapman's Geeks, Mops, and Sociopaths also comes to mind.)
This point makes me more optimistic about the ability of humanity to pause frontier AI progress in the near future.
3-5 years ago AI models were tool-shaped or chatbot-shaped so it was hard to convince normal people that AI was an existential risk. One explanation why is that normal people tend to see the present state of the technology (e.g. 2022 ChatGPT) whereas LW-type people see the long-term trajectory of the technology which is useful but difficult to convey quickly.
But now with the advent of agentic AI and AI swarms, AI models seem like people or groups of people and now it's easy for normal people to see the risks of AI if they're already worried about uncontrolled immigration because the concepts are similar.
Advanced AI is basically the embodiment of immigration as envisioned in the conservative nightmare:
Whether or not you think this is a good description of the situation with foreign humans joining your country, it is a good description of the likely AI to come, and it’s even worse than imagined: