One thing I think is weird about all of this, though, is that there’s a definite movement that is at least adjacent to the tech world if not within it, where people take this stuff (edit: by this stuff I mean the critique you outline in your OP) way more seriously. Jasmine is part of that movement, imho!
Is there anything like that in anthropology? Genuinely curious here.
My impression is that if you’re critical of this mentality, then you’re way more likely to be in tech or tech-adjacent, than you are to be anywhere near academia - but maybe I’m incorrect about this?
Ironically in actual literal anthropology, I think people like Joe Henrich have been leading a bit of a charge against the existing normative mode, while trying to not self-parodize into Anti Academia Brand Evangelists. Part of what is distinct about the spaces is that I think academia has been having successive iterations of this kind of conversation since at least the Bloom era, and arguably the Buckley era, so the ‘grooves of the record’ are a bit easier to fall into camp-wise, I think. But also that’s very much not my space.
I agree the whole kind of movement Jasmine and many others are a part of and embodying matters a whole lot here.
I think you're making a decent argument, but I (in a similar pace culturally re AI as you), would disagree on two fronts:
On regular folks gumming stuff up for thymotic reasons, Jasmine's article didn't read that way at all to me. Yes, a person SAID some stuff along those lines, but the more widespread main thrust that I got was "money was promised in the past and it never paid off, ie it wasn't contractually bound at all," along with a tiny bit of your "people hate tech / tech oligarchs and this is a way to stick it to the man" point. You know, instead of just literally not infinite scrolling for the 7-9 hours a day people put into phones now.
But because all the hyperscalars have more money than god to deploy, the first problem, which I see as the dominant one, can be easily solved with iron clad contracts with prepayments and forced commitments to their tax or education budgets or whatever. And I'd bet that solves 80% of them, 95% at a high enough benefits level. The main problem is a credible commitment mechanism, and precommitment / prepayment works for that. And at the scale they're deploying capital, a prepayment of a few tens of millions is fine if that place actually has grid interconnect capacity, and would still be well worth it economically. Data centers make absolutely absurd margins, and that's before AI inference grew from 40% margins in 2025 to 70-80% margins in 2026.
> It’s “we need to barrel forward so fast that there can’t be any especially meaningful inhibition to what we’re doing.”
I will argue that this is just straightforwardly the right move, because not doing so, or trying to be "democratic" or trying to elevate this to government attention right now, straightforwardly increases existential risk millionsfold.
Trying to invoke "democracy" leads to vastly more risk and even more danger for all of us, including the regular folk in flyover country.
The people actively advocating a path like this right now, the Fronter AI Institute / Plan A people, Scott Alexander, Zvi Moshowitz, the slowdown letter people, are frantically facetiming Sauron on one of those crystal ball things (palantirs for the nerds out there, or the nominative determinists), yelling “Sauron, Sauron, we have a gigantic ring of power, and it keeps getting more powerful, please come and help us coordinate with the Chinese hobbits…wow, he’s already here! Yes, thank you Sauron, could you please tell the Chinese hobbits…wait, he’s picking it up. No! He’s putting it on! Nooooo!”
This is because "democracy" doesn't mean "the people decide," it means "the government decides," and the ONLY power centers in the US government that can plausibly even pretend to have people able to understand anything at all about frontier AI models are the NSA and the military. And that's simply how government works - they go to the biggest power centers relevant to the thing in question, and those power centers get the juice and the ability to shape all the narrative and decisions from there. And once the military + NSA is directing frontier AI development and directions, it increases existential risk for us all millionsfold, because obviously China will do the same, then the race is going to be even MORE on, and basically unstoppable.
And now this unstoppable race's ends and means have been fundamentally perverted in "spy on everyone" and "kill everyone" directions. NOT what we want, as humanity overall!
And "the people" will not be able to change or influence that, because again, it's about existing power centers in the government. I mean, tell me how "democracy" is going to prevent that, because I dearly want that path to exist, and I simply think it doesn't.
"Trying to be 'democratic'...increases existential risk millionfold."
I genuinely can't believe how well this proves Zak's whole point. Why the hell would I, or anyone else, dream of trusting the judgement of someone who says "these dumbass hicks just don't get it"? You keep insulting normal people in lieu of actual engagement, then give a *Pikachu shocked face* when we say "go fuck yourself." Jesus, I'm being polarized against AI by the day, and I should be a supporter!
Yeah this is kind of the whole dynamic at play, as you say - it’s one reason *the very parameters of the conversation* tend to be so profoundly alienating.
I hate to be abrasive; it's not my nature! But is it a crime to say "let's hold up for a moment?" Particularly when AI companies seemingly aren't looking out for the interests of me, my community, or even *humanity writ large*.
I don't believe that the world will be changed by AI (to extinction or utopia) when barely a single person in that world acts as if it's true. I would take out loans, max credit cards, and live as if the world was ending if I thought the world was ending/drastically changing forever. Nobody close to AI is living that way, so it smells like BS. Yet they say it, incessantly, so I have to take it somewhat seriously.
I, personally, would love to contribute to AI safety. But I can't think of a way to effect change other than to protest a data center. If one comes to my community, I'll oppose it. Not because it's the Best Decision, but because it's *the only thing I can do* to maybe make AI companies listen or care.
Totally - and I think this attitude makes a lot of sense! Not every option is equally possible, equally legible, equally effective, or equally satisfying. Sometimes the thing to do is the thing that's most straightforward!
PauseCon seems laudable, but I'm not sure how applicable it will be to me as an American. If they put up videos related to organizing, lobbying, educating, etc, I will watch those!
I think as US citizens, this is our responsibility more than almost anyone, since so much AI development is happening in our country.
I myself am chronically ill, so unable to attend conferences, but I do try to put in hours raising awareness online. Lots of bad reasoning out there to dismantle. Luckily, I'm undefeated in internet arguments.
Lots here; agree with some. In particular I do agree that state capacity is integral to all AI ‘alignment’ conversations for this reason. But I don’t think the contracting solution solves the trust problem - I don’t think most people have a model for like “what is the actual rate of followed-through-upon promises” (certainly I saw the range of assertions about this in re: Colossus).
And I just don’t think “hope to subvert mechanics of governance” ever lasts.
I'm interested in your take on how much of this even *is* solvable with better behavior from the hyperscalers. I think it is more than 0 but probably a lot less than 100%.
I agree with your broad point in the article that it is both counterproductive and silly to ascribe all opposition to you/your group (whether its the valley or academia etc) to the fault of your audience. However, I also think it's true that social media and the conspiratorialism and rage generated by the algorithmic feeds makes rational discussions on anything borderline impossible, no matter how willing you are to recognize your own contribution to the problem at hand.
For full disclosure I am pro-AI and pro-data-center (not just for AI but also because data centers are the entirety of the internet), but I also believe that communities should not be incentivizing development of data centers with tax abatements and sweetheart deals when there is such a gigantic need in the industry for more compute.
Yeah - I agree with all of that. I think the burr in my heels is that we literally can’t know how much trust can be engendered without the securing of such trust first being named as an actual goal, as an actual obligation. It feels almost quaint now to think in terms of duty but it was not so long ago that the sense of *duty* conferred by AT&T’s monopoly status gave us Bell Labs, which gave us all of this. We have time yet to turn towards trust.
Tech Hyperscalers Yield Maximal Opposition Surface
😆😆😆
Elon wants to put the data centers up in space so community meetings can't block them:
https://www.a16z.news/p/spacex-and-the-sentient-sun
That way when the rogue AI gets going, it will be totally infeasible to unplug it...
Honestly that’s a meaningful driver for sure. Of course, that’s why the moon should become the 51st state ;)
One thing I think is weird about all of this, though, is that there’s a definite movement that is at least adjacent to the tech world if not within it, where people take this stuff (edit: by this stuff I mean the critique you outline in your OP) way more seriously. Jasmine is part of that movement, imho!
Is there anything like that in anthropology? Genuinely curious here.
My impression is that if you’re critical of this mentality, then you’re way more likely to be in tech or tech-adjacent, than you are to be anywhere near academia - but maybe I’m incorrect about this?
Ironically in actual literal anthropology, I think people like Joe Henrich have been leading a bit of a charge against the existing normative mode, while trying to not self-parodize into Anti Academia Brand Evangelists. Part of what is distinct about the spaces is that I think academia has been having successive iterations of this kind of conversation since at least the Bloom era, and arguably the Buckley era, so the ‘grooves of the record’ are a bit easier to fall into camp-wise, I think. But also that’s very much not my space.
I agree the whole kind of movement Jasmine and many others are a part of and embodying matters a whole lot here.
I think you're making a decent argument, but I (in a similar pace culturally re AI as you), would disagree on two fronts:
On regular folks gumming stuff up for thymotic reasons, Jasmine's article didn't read that way at all to me. Yes, a person SAID some stuff along those lines, but the more widespread main thrust that I got was "money was promised in the past and it never paid off, ie it wasn't contractually bound at all," along with a tiny bit of your "people hate tech / tech oligarchs and this is a way to stick it to the man" point. You know, instead of just literally not infinite scrolling for the 7-9 hours a day people put into phones now.
But because all the hyperscalars have more money than god to deploy, the first problem, which I see as the dominant one, can be easily solved with iron clad contracts with prepayments and forced commitments to their tax or education budgets or whatever. And I'd bet that solves 80% of them, 95% at a high enough benefits level. The main problem is a credible commitment mechanism, and precommitment / prepayment works for that. And at the scale they're deploying capital, a prepayment of a few tens of millions is fine if that place actually has grid interconnect capacity, and would still be well worth it economically. Data centers make absolutely absurd margins, and that's before AI inference grew from 40% margins in 2025 to 70-80% margins in 2026.
> It’s “we need to barrel forward so fast that there can’t be any especially meaningful inhibition to what we’re doing.”
I will argue that this is just straightforwardly the right move, because not doing so, or trying to be "democratic" or trying to elevate this to government attention right now, straightforwardly increases existential risk millionsfold.
Trying to invoke "democracy" leads to vastly more risk and even more danger for all of us, including the regular folk in flyover country.
The people actively advocating a path like this right now, the Fronter AI Institute / Plan A people, Scott Alexander, Zvi Moshowitz, the slowdown letter people, are frantically facetiming Sauron on one of those crystal ball things (palantirs for the nerds out there, or the nominative determinists), yelling “Sauron, Sauron, we have a gigantic ring of power, and it keeps getting more powerful, please come and help us coordinate with the Chinese hobbits…wow, he’s already here! Yes, thank you Sauron, could you please tell the Chinese hobbits…wait, he’s picking it up. No! He’s putting it on! Nooooo!”
This is because "democracy" doesn't mean "the people decide," it means "the government decides," and the ONLY power centers in the US government that can plausibly even pretend to have people able to understand anything at all about frontier AI models are the NSA and the military. And that's simply how government works - they go to the biggest power centers relevant to the thing in question, and those power centers get the juice and the ability to shape all the narrative and decisions from there. And once the military + NSA is directing frontier AI development and directions, it increases existential risk for us all millionsfold, because obviously China will do the same, then the race is going to be even MORE on, and basically unstoppable.
And now this unstoppable race's ends and means have been fundamentally perverted in "spy on everyone" and "kill everyone" directions. NOT what we want, as humanity overall!
And "the people" will not be able to change or influence that, because again, it's about existing power centers in the government. I mean, tell me how "democracy" is going to prevent that, because I dearly want that path to exist, and I simply think it doesn't.
I wrote about this in a post recently.
https://performativebafflement.substack.com/p/contra-plan-a-scott-and-others-lets
"Trying to be 'democratic'...increases existential risk millionfold."
I genuinely can't believe how well this proves Zak's whole point. Why the hell would I, or anyone else, dream of trusting the judgement of someone who says "these dumbass hicks just don't get it"? You keep insulting normal people in lieu of actual engagement, then give a *Pikachu shocked face* when we say "go fuck yourself." Jesus, I'm being polarized against AI by the day, and I should be a supporter!
Yeah this is kind of the whole dynamic at play, as you say - it’s one reason *the very parameters of the conversation* tend to be so profoundly alienating.
I hate to be abrasive; it's not my nature! But is it a crime to say "let's hold up for a moment?" Particularly when AI companies seemingly aren't looking out for the interests of me, my community, or even *humanity writ large*.
I don't believe that the world will be changed by AI (to extinction or utopia) when barely a single person in that world acts as if it's true. I would take out loans, max credit cards, and live as if the world was ending if I thought the world was ending/drastically changing forever. Nobody close to AI is living that way, so it smells like BS. Yet they say it, incessantly, so I have to take it somewhat seriously.
I, personally, would love to contribute to AI safety. But I can't think of a way to effect change other than to protest a data center. If one comes to my community, I'll oppose it. Not because it's the Best Decision, but because it's *the only thing I can do* to maybe make AI companies listen or care.
Totally - and I think this attitude makes a lot of sense! Not every option is equally possible, equally legible, equally effective, or equally satisfying. Sometimes the thing to do is the thing that's most straightforward!
Josh and your points are the Midwestern mentaility in a nutshell.
"I can't think of a way to effect change other than to protest a data center."
This might be interesting: https://pausecon.org/
PauseCon seems laudable, but I'm not sure how applicable it will be to me as an American. If they put up videos related to organizing, lobbying, educating, etc, I will watch those!
Hadn’t actually heard about this - thanks for the share!
Looks like there's another PauseCon in DC in mid-September. See banner at the top of this page:
https://www.pauseai-us.org/
I think as US citizens, this is our responsibility more than almost anyone, since so much AI development is happening in our country.
I myself am chronically ill, so unable to attend conferences, but I do try to put in hours raising awareness online. Lots of bad reasoning out there to dismantle. Luckily, I'm undefeated in internet arguments.
Lots here; agree with some. In particular I do agree that state capacity is integral to all AI ‘alignment’ conversations for this reason. But I don’t think the contracting solution solves the trust problem - I don’t think most people have a model for like “what is the actual rate of followed-through-upon promises” (certainly I saw the range of assertions about this in re: Colossus).
And I just don’t think “hope to subvert mechanics of governance” ever lasts.
I'm interested in your take on how much of this even *is* solvable with better behavior from the hyperscalers. I think it is more than 0 but probably a lot less than 100%.
I agree with your broad point in the article that it is both counterproductive and silly to ascribe all opposition to you/your group (whether its the valley or academia etc) to the fault of your audience. However, I also think it's true that social media and the conspiratorialism and rage generated by the algorithmic feeds makes rational discussions on anything borderline impossible, no matter how willing you are to recognize your own contribution to the problem at hand.
For full disclosure I am pro-AI and pro-data-center (not just for AI but also because data centers are the entirety of the internet), but I also believe that communities should not be incentivizing development of data centers with tax abatements and sweetheart deals when there is such a gigantic need in the industry for more compute.
Yeah - I agree with all of that. I think the burr in my heels is that we literally can’t know how much trust can be engendered without the securing of such trust first being named as an actual goal, as an actual obligation. It feels almost quaint now to think in terms of duty but it was not so long ago that the sense of *duty* conferred by AT&T’s monopoly status gave us Bell Labs, which gave us all of this. We have time yet to turn towards trust.