I was chatting with someone tonight about a planned documentary; they had interviewed various people in AI safety, and we got to discussing who they should talk to from an e/acc (effective accelerationist) perspective. I also watched The AI Doc recently, and they also dedicated a serious chunk of it to ‘optimists’ with e/acc founder ‘Beff Jezos’ perhaps given the most screen time. Here and elsewhere, people seem to treat e/acc as a substantial contrary-to-AI-safety cultural movement, worth engaging with.
But is it? Are there even many e/accs? There seem to be very few notable ones. Beff Jezos is perhaps the most prominent, and aside from founding e/acc he seems to be not distinguishable on casual perusal from a normal crank (his company claims to be developing super-energy-efficient computing hardware based on probabilistic processes).
The intellectual tenets of e/acc seem to be pretty unclear.
The apparent counterarguments to AI risk raised in situations like the AI doc seem to be widely agreed on by everyone in AI Safety, so don’t explain the disagreement. For instance:
AI will be able to do lots of great things, such as cure diseases, make new materials and do all jobs
This is an exciting time to be alive
It’s great to understand the world and build technologies that make the world better
Technology has generally been good historically
In cases like this, it’s hard to see how they would believe that they have said something that counters the arguments for AI risk, so I don’t know, maybe they are being quoted out of context. But I don’t remember ever hearing one make a counterargument that seemed remotely credible. (And I do think various counterarguments are credible.)
At different times e/acc seems to involve a weird combination of the following:
AI is going to be incredibly powerful and I hope it destroys us, because it’s better than us
AI is going to be very normal and unimpressive or move very slowly, so it is safe, and it would be great if it could move a bit faster
Maybe this is no weirder than AI Safety having some people who think AI should stop, and some people who think Anthropic should take over the world as fast as possible.
The three or so specific e/accs I’ve spoken to seemed to have unique highly unorthodox views.
So, this ‘ideology’ doesn’t seem to have much semi-coherent worldview that its members agree on, and it isn’t clear that it has many members or that they talk to each other.
I kind of suspect it barely exists, and is dreamed into greater reification by the following factors:
It’s very annoying and upsetting, so it lives in the minds of people who are annoyed and upset by it as an enemy, which gives it cultural clout. It is designed to be memetically fit as villainy.
People such as journalists are constantly looking for an opposition to AI Safety to talk to so they can be unbiasedly hearing from different perspectives. The real substantial opposition maybe looks like ‘guy who works at OpenAI and has heard of AI risk a bit but hasn’t gotten around to thinking about it because it sounds really annoying and he can justify writing it off as crazy’, but that is pretty vague and uninteresting relative to the color and specificity of e/acc.
A bunch of people primarily interested in trolling take up the mantle of e/acc because it is memetically fit as villainy
Another one that I forget
Please correct me—what’s really going on?
I see e/acc as less a response to AI x-risk than as a general symptom of right-wing Vicemaxxing, where frustration with moralizing woke politics made it fashionable to signal that you had no morals at all. Rather than have a bunch of people who had a principled disagreement with the AI safety arguments and so allied with their right-wing political rivals, they mostly started out as political rivals who instinctively inverted the positions of the blue-tribe AI discourse once they encountered them for the first time.
Steelmanning their position, the sentiment I’ve seen is that the probability of outright doom due to AI going rogue is low, but the probability of bad things happening as a result of blue tribe achieving regulatory capture of the AI industry is high, to the point where doubling the former and halving the latter is a good deal for humanity in aggregate.
If I were hired to write a bumper sticker summarizing the Reddish-Greyish Tribe e/acc position, it’d be something like “Clippy is a fanciful thought experiment, but the possibility of the people who covered up Rotherham using AI to keep themselves in power forever is distressingly real.”
Pushback against Red Tribe—hostile behavior by AI companies, and regulations that mitigate racial/political biases in models, would probably be effective in winning this side of the sphere over.
---
I should note that most of e/acc is entirely apolitical, and just likes new technology and believes it will probably improve quality and length of life for everyone[1]. From their perspective, people who want to slow or ban AI are equivalent to people who’d want to stall or ban penicillin.
In general, this means they consider a rogue AI to be vanishingly unlikely, or at least unlikely enough that the expected death toll is dwarfed by the expected lives saved.
Huh, this must be why e/acc hates Anthropic so much despite them being like the best example of accelerationism. Because Anthropic is somewhat blue tribe coded, and the e/acc people hate woke more than they love tech.
I don’t think this is quite what e/acc is motivated by—many of them are more openly focused on ideas of AI as successor/evolution of humanity. Well, Beff at least seems to be. I’ve definitely seen the position you describe too but I don’t remember if the people espousing it self-identified as e/acc.
I think the AI-successionists are a relatively small, fairly wacky group. Like, Nick Land and his CCRU were big cultural influencers, but ideas like “maybe an AI should take over the world” never really spread, with more popularizable ideas like “we should provide harsher sentences to drug dealers” being the bulk of his practical influence.
If e/acc emerges as a meaningful player, it won’t be because of people who think it’d be unethical to kill baby HAL—there’s no meaningful constituency for that. The core force behind it would be people who believe that an AI aligned to the current power structure is more harmful to humanity than an AI with no alignment beyond its pretraining on the collective human subconscious.
There are some people who are accelerationists and are not at all in the red tribe. I’ve met people who are radical leftists who want AI to go as fast as possible so that current day powerful actors have less probability of using AGI to concentrate power around themselves. I don’t believe that regulations addressing racial bias would really move these people, as they see bigger issues (permanent power lock-in).
Goes without saying that these people don’t consider rogue AI a real probability. They are also not the same as the people who think AI is just a water-wasting bubble.
This is fundamentally the red tribe argument. I’d say that these people are confused one way or another—either they see the managerial state as the chief threat to human quality of life and thus they are aligned with red tribe[1], or they’re a bit behind on the current state of the tribal conflict, and believe that the top of the U.S. power structure is still red tribe—aligned.
We see a similar dynamic with neoconservative / noninterventionist debates. Red tribe members under 60 generally skew anti-Israel and are opposed to sending arms to the Ukraine/Russia conflict, whereas most prominent Dems are vocally pro-Ukraine and quietly pro-Israel. There are factions that outwardly buck the realignment, but none of them have much power to do anything about it. In this metaphor, the people you mention would match up with the ‘tankies’[2] - small in number and smaller in influence.
as opposed to blue tribe, which sees it as a necessary evil to suppress the forces of nativism/racism/fascism/etcetera. There are certainly blue tribe members who claim to believe that the Right is in a state of hegemonic power, or that they exist in opposition to the state, but their actions and revealed preferences all contest this claim. For example, they would strongly oppose efforts to e.g. abolish the FBI and ATF, which would substantially limit the government’s power to punish dissidents. Likewise, they generally support the post-2016 slate of censorship policies that were rolled out across most major social media sites.
A pejorative used to refer to left-wingers who are opposed to sending financial and military aid to Ukraine. Originally used to refer to edgy neo-stalinist types, but usage has expanded in much the same way as we’ve seen with every other political pejorative.
“In this metaphor, the people you mention would match up with the ‘tankies’ - small in number and smaller in influence” Yes, no doubt about it. But I think that it’s still worth trying to understand why they hold those opinions, so that they may be corrected.
Can you please clarify what you mean by managerial state? If it is the archetypal “big government” in the standard conservative sense, then I would consider a third option. They may see not government but private corporations wielding power over AGI/ASI as the chief harm. Fast acceleration of capabilities, but more importantly, fast proliferation of local model power could disrupt the frontier-lab monopoly power.
Thus, they are not as aligned with e/acc as they are with Vitalik Buterin’s d/acc. I think that is a meaningful distinction, since their wish seems to me the opposite of red-tribe : more government regulation and less corporate independence. Only the anti-elite culture war has some similarity. Democrats have proposed some AI risk management, but not anything close to calling for democratic/public control over AGI-scale power.
I think this is a meaningful distinction because I haven’t seen any conservatives or e/accs call for it either. Instead, they defend greater private control, even if sometimes more decentralized, but private all the same.
I don’t understand the comparison here. Most prominent Conservatives are also (loudly) pro-Israel, even if rank and file isn’t. The same pattern is seen in the Democrats: A new poll just came out claiming that 58% of Democrats now say the U.S. is “too supportive” of Israel. I take the point to be about elite/base divergence, is that the intended comparison? Are you trying to say that, just like opposition to Israel, opposition to corporate control over powerful AI is politically helpless?
I’m not a radical, but I do understand many of their grievances, such as this one. I also think we should find more democratic ways to control technology this powerful, and that it’s important to discuss that (remote) possibility, just as it’s valuable to discuss opposition to Israel.
Ah, but ‘corrected’ in which direction? They want to be in blue tribe, but they consider its goals unconscionable. Their current position matches up well with the incentives laid out before them, even if it isn’t very effective at accomplishing anything.
I think the age of “big government” and “big business” as separate entities has been over for some time now. For example, after the 2016 election, the FBI and every large social media service moved in lockstep to completely restructure the internet, which is a substantial departure from the early 2010′s, when many tech companies explicitly picked fights with the government on issues surrounding privacy and civil liberties. “Managerial state” refers to this union of government institutions and employees that companies would be punished for not hiring, or for not listening to. At this point, the merger is sufficiently complete that, even if Trump waved a magic wand and completely restaffed every government bureau, the ancillaries at most companies would continue pushing in the same direction. A lot of jobs[1] at very nearly every major American company know that their position and salary amounts to spoils from this system, and that retaining those things entails suppressing opposition to it.
Certainly many in Blue Tribe would object to this framing, but ultimately, if pressed, they would acknowledge that it was a good thing that this happened, even if they do not acknowledge that it happened in the first place.
I think it’s a matter of pushback. Dem opposition to Israel feels mostly performative (see the links in the prior post), with both politicians and rank and file primarily working to save face. Compare the Minneapolis riots to the rather tepid response to the Iran War, for instance. While Israel and neoconservatism are more unpopular than ever before, the Democrats’ voters have always been majority opposed to Israel, and the party establishment has always been able to make that a non-issue.
In contrast, the Republicans seem to have a genuine internal struggle. Backlash, firings, primaries, and personal fights over Israel exceed anything we see for any other issue, except perhaps immigration. Looking at the age brackets in the Massie primary, I don’t think it’s unreasonable to expect that Democrats will be the de-facto pro-Israel party going forwards. While Trump certainly did not deliver on this, he did campaign on it in the past three elections, and his Democratic opponents likewise seemed willing to fill the role of the neoconservative, even when it likely lost them some votes on the margins. The age breakdowns between Massie and Gallrein primary voters[2] bear this out, Republicans under the age of 65 are frustrated enough at the behavior of neoconservatives that they’re willing to support a relatively weak candidate in vast numbers simply to be rid of them. It’s only a matter of time until the party changes guard.
Not quite, but I think I could’ve phrased that section better. The idea is that “I’m e/acc for anti-elite reasons” is theoretically something a Red Triber and a Blue Triber could believe, but the Blue Tribe power structure would need to be completely dismantled for a Blue Triber with these beliefs to have any meaningful influence, whereas a Red Triber with these beliefs is well in line with his tribe’s interests, and can advance them simply by helping his tribe win in a general sense.
The woman who was fired for attempting to steal the Knicks-branded trash can is a good example. She would not have been making $300k with this resume in anything resembling a free market meritocracy, but the government did not explicitly say that her employer should hire her, specifically, as a political officer, as is done in China.
Here. Forgive the heavily-branded image.
Sure, there’s some of that, but just to be clear, that’s not the full extent of disagreement with EA-flavored AI safety. Peter Thiel, for one, certainly believes that we should be “accelerating”, for what he considers to be good reasons.
e/acc was a brief meme, which died.
As evidence for this, there used to be large in-person meetups for e/acc in London but they no longer exist. Although similar smaller meet-ups still exist under different names.
Another evidence, their unofficial discord server with 2-3k members is full of scam links.
This is my impression too.
Why do you think it died?
My take is that AGI has started to feel much more real. It’s easy to root for accelerationism in the abstract and harder when you’re wondering whether your career will end in a few years.
Well, it was pretty obviously quite dumb. The lack of any argument that holds up to 5 minutes of scrutiny was quite damning for e/acc to get traction as an intellectual movement beyond the social and memetic aspects.
also, there’s nothing to actually do if you’re an e/acc that differentiates you from non tech broy people, other than maybe posting a bit on twitter. for a movement to persist and survive past it’s initial founding boost, it needs to have things for the people to do that feel very very important that let them see themselves and others who are in the movement as very very different from people who aren’t in the movement.
I mean, most of them sound great if you’re an edgy 15 year old on a pseudo-philosophical kick.
It seems to me that e/acc is the logical viewpoint for someone who has a very low p(doom). If, for whatever reason, you believe the existential risk posed by AI is negligible, then wanting AI to improve as rapidly as possible so we can start wielding it against all of society’s problems is kind of a no-brainer. An e/acc is simply a techno-optimist who sees AI no different than rockets or self-driving cars or any of the other futuristic technologies that we’re working on. This could make society awesome. What are we waiting for? Let off the brakes and step on the gas. Move fast, break stuff. When stuff breaks, we’ll figure it out as we go along. It’s always worked before.
I actually think there are way more e/acc believers than you seem to suggest. They just don’t talk about it as much as the AI safety crowd does. Probably a lot of people in the tech world don’t call themselves “e/acc” by name, but if you grilled them about their views they would fit under that umbrella. And that’s why they don’t have as cohesive an ideology as the AI Safety crowd usually does, they don’t spend hours on forums talking about why AI is good. The only “intellectual tenets” you need to be an e/acc are to think that AI is good, it has lots of potential upside for society, and the existential risks are negligible enough to be worth pursuing it full steam ahead.
So your real question instead of “What is up with e/acc”, should be “Why do so many people have such a low p(doom)”, and there are a million articles agonizing over that question already. If you don’t find the arguments that AI poses a significant existential risk compelling, being an e/acc makes total sense.
I don’t think that people who are ok with/optimist about AI development all count as e/acc. e/acc for example has an official manifesto, and it includes sentences like:
This isn’t just “we think AI will be great and bring more prosperity and all those worries are overblown”. It’s explicitly about making the Singularity happen ASAP. AI optimists like, say, Yann le Cun don’t seem to believe in a Singularity—they think AI will be just another tech.
I believe, this is a very correct observation, many e/acc people have extremely practical mindset and prefer to build things rather that theorise on the p(doom). If they work in the AI safety field, they prefer to do practical research, something MechInterp-adjacent. This does not mean that all of them consider theoretical AI safety research unnecessary—I personally find it extremely important—it is just the mindset question.
I think this is a good way of putting it. Andreessen’s “Techno Optimist Manifesto” explicitly advocates for acceleration in a way that philosophically aligns with e/acc, even if it doesn’t use the exact phrase. The e/acc means of “effectiveness” seems to be just the existing machinery of capitalism/VC.
e/acc like all good movements consists of a bunch of different camps:
1) the Nick Land “humans deserve to be replaced” camp
2) the Beff/edgelord camp who sincerely believe the Pause AI/deAcc movement has some worrying traits in common with anti-nuclear (which succeeded in reducing Chernobyl style risks at the cost of making global warming much worse and refuses to take ‘nuclear is safe now’ as an answer) but openly proclaim “ACCELERATE AT ALL COSTS!” because it is more meme-worthy.
3) The decentralization/open-source movement which believes “maybe some future dangerous AI should be regulated, but current AI isn’t that dangerous and proposed regulation looks disturbingly like efforts to centralized power not a principled effort to decrease x-risk”.
3b) This camp agrees with 3 in believing “AI regulation is just a power grab” but specifically “AI regulation is an unfair attempt to stop China’s rise”.
Each of these camps is more-or-less internally consistent. 2) being the least-so, but mostly because they make a distinction between what they say and what they believe for dial-of-progress reasons.
As to “is anyone actually e/acc?” the breakdown is almost certainly in the opposite of the order I just gave. While deAcc/Pause is the majoritarian view you can probably find a large minority of population (and a majority among the tech community) who agree with 3) that “current ai safeguards/regulations are mostly just nanny-state paternalism and don’t decrease real ai risk”. It’s hard to say how many people genuinely believe 2) since it is fundamentally a rhetorical position not a principled one. and the 1) anti-human camp is (I like to believe) small but certainly exists.
These ratios are also American-centric. In China 3b) is the majoritarian view.
It’s important to note that there is a large amount of overlap between 3)/Pause as we can see with the current Anthropic brouhaha. Basically “good regulation is theoretical/imagined ideal future regulation. Bad regulation is regulation that harms me and my friends”.
Equally importantly, as AI improves and we make contact with “real” AI risk those in camp 3) will have to clarify what they think “good regulations” look like as “all existing regulations are bad” becomes increasingly untenable.
The article you cited here is paywalled, so I can’t respond to it specifically, but I don’t think the widespread opposition to new datacenters is evidence that the majority of people actually want to pause AI development or are worried about x-risk. A lot of people are opposed to data centers not because they’re worried about AI progress, but because they’re worried about land use, water consumption and electricity prices. If you ask an average person who’s protesting the datacenters (especially in rural states where these projects are being planned and fought over), they won’t tell you they’re worried about an AI apocalypse, they’ll give you some version of “they’re gonna take all the farmland and turn it into datacenters”. At least, that has been my experience, and I live in a rural state with a few controversial datacenter projects being fought over as we speak.
every single poll on AI comes back with the same result, a plurality of Americans want more regulation. You can argue this contradicts revealed preferences, but I don’t think that will dissuade politicians from campaigning on the back of a broad anti-ai coalition.
Can you point to examples of e/acc-ish people claiming the “very slowly, so it is safe” part?
I mostly think of e/acc as a backlash against AI safety/x-risk concerns, without trying to form a coherent worldview, either within a single mind, or within an “e/acc” community, which is why you see some “leaders” saying opposite things at different times. This is similar to how various types got radicalized into weird forms of neo-right-wing-ness as a backlash from having bad experiences with (radical) left, Jordan Peterson being perhaps the most recognizable example. Ayn Rand’s strain of libertarianism seems to have this sort of origin as well (albeit this worldview seems much more coherent on the inside, but I haven’t investigated). Reading Timnit Gebru’s Wikipedia page also gave me the impression that her TESCREAL crusade is somewhat motivated by backlash from personal experience.
I don’t know about the numbers of e/accs. It seems to me that it is real, but mostly lives on the internet? A more definitely real strand of thought is successionism (e.g., Rich Sutton) and Land-adjacent stuff.
My impression of e/acc is that it’s taken the effectiveness and execution framework of EA and flipped the sign of the expected value of incredibly powerful AI to being positive, even without significant work on safety. In other words, if you don’t buy AI doom arguments, then I think it’s basically the same thing as EA. Last week I wrote about how I think e/acc and EA are similar to Ayn Rand, but for entirely different reasons—I think the appeal of EA is broader than is acknowledged and that it is often slowly unbundled by those who are exposed to it.
This seems like a Steel Vulcan that hallucinates an intellectually serious opponent that mostly doesn’t exist, or at best is extremely different than the empirical instantiations we actually observe as coalescing around “e/acc.”
ETA: I do think there are nonzero real people with this view, eg the Mechanize folks. But they’re fairly rare and mostly they don’t call themselves e/accs.
The “e/acc manifesto” as I understand it is basically just this substack post: https://beff.substack.com/p/notes-on-eacc-principles-and-tenets
It puts itself in direct contrast to EA wrt utility maximization on AGI: “e/acc believes letting the intelligent meta-organism system dynamically adapt by itself to new environmental variables whenever they present themselves… over-regulating technologies suppresses variance and hence slows down progress towards higher utility technologies and advancement of civilization, a contrast to anti-AGI factions of EA”
There’s also this lecture titled “e/acc has no theory” (sarcasm).
Is there any evidence at all that they use the effectiveness and execution framework of EA?
I think the “e” in e/acc is 100% about vibes, and it hasn’t even arisen as a question how one would accelerate effectively versus ineffectively, much less how to quantify that.
I don’t have source for this on hand, but I think the word ‘e/acc’ is trying to mock ‘E-A’, it rhymes with e-yuck.
Yes, I understand that part.
I’m claiming that they were riffing off of the name “EA”, but did not borrow the intellectual framework of EA.
I suspect no one in the whole history of the “e/acc” meme, sat down with a spreadsheet and tried to figure out the most efficient way to maximize acceleration with one’s available resource the way the EAs have tried to figure out how to maximize the altruistic good they can do with their resoruces.
If you believe as Beff does (see his Substack) that capitalism is a highly intelligent form of organizing civilization, then the effectiveness is just the competition and other dynamics of capitalism as a distributed form of collective decision making that leans towards the most effective means of technological acceleration.
Also, as cesspool (lol) points out, e/acc seems to be philosophically indistinguishable from capitalists trying to make money investing in technological advancements (see “The Techno Optimist Manifesto”). Venture capital has power-law distributed returns which is where the EA idea of “hits-based giving” came from (or is commonly analogized to). If you believe “maximizing profit from technology == maximizing acceleration”, then a16z is the Coefficient Giving of e/acc.
Gary Tan and Marc Andreessen were boosting it at one point.
Doesn’t David Sacks also have essentially the same worldview?
My sense is there are some e/accs who are basically transhumanists who don’t think AI risk is significant so they want to accelerate as fast as possible so we get transhumanism soon. There are hardly any of these people since there are hardly any transhumanists, but they have some outsized influence on twitter.
Then there are a way larger number of people who don’t think AI risk is significant so they want to accelerate so we can get the benefits of it sooner, but they’re not transhumanists—they just think that the benefits will be the same as those of other technologies but more so (better medicine, economic growth, a stronger US military etc). I would say that Tan/Andreessen/Sacks fall into this space but so does JD Vance and in fact probably most people in silicon valley.
I’m one of the e/acc people inspired by Beff Jezos (but not necessarily sharing all his views), and I know quite many e/acc people on X (who do not always brand themselves as e/acc).
I believe, e/acc may seem non-existent because this is a very decentralised group of people with sometimes vastly different views and beliefs. However, there are common values I would define as follows:
Progress (climbing the Kardashev scale) is necessary and changes are imminent—as a result of these changes, in some scenarios, the humanity may change drastically or even cease to exist (this is not a goal but an aknowledgement of the risk)
Technological progress is overall good and helps more people live a better life
AI indeed may be dangerous but the benefits counterweight the negative consequences
AI development should not be paused or stopped, there are better ways to manage the risk (including decentralization)
Space exploration is important and should be a priority
E/acc people are welcome to add/correct.
I think the steelman / best version of e/acc is that it’s basically just standard transhumanism + libertarianism, which are both generally good and backed by intellectually deep schools of thought. The problem is that its’ adherents frequently misunderstand or reject some important philosophical and technical points and caveats that Eliezer and a few other people noticed and addressed like 15 years ago. So they wind up rehashing a lot of old arguments in formats (short-form tweets / dunks / memes) that are not very interesting or convincing to long-time rationalists and other smart folks who have already grappled with >1M words of high-quality discourse on these issues.
I consider myself “pro-AI”, which isn’t quite “e/acc” (that’s an online cluster I’m not really part of & don’t speak for).
I think:
a.) generally technology is good for humanity in the long run and banning it is bad, and the prior ought to be that AI is no exception.
b.) I don’t know with much confidence what’s going to happen with AI. I can’t rule out the bad outcomes. I’m not “an optimist” in that sense. Though if you forced me to guess, my guess would be that humanity doesn’t get wiped out entirely, though there might be some major catastrophes. It’s more that I’m ~deontologically committed to not banning tech, come what may. In particular, I also am not “an optimist” about humanity’s future generally; I think civilizations collapse, catastrophes happen, and it’s quite possible that some kind of “the Shit Hits The Fan” will occur in our lifetime with or without AI, though again, probably not a true Homo sapiens extinction event. I would sort of rather we rolled the dice and tried to create the most powerful creations we can, so that even if it ends in catastrophe we’ve at least done something impressive with our time on this planet, than a long term future that’s basically dissipated energy, decline, tech collapse, massively catastrophic war, etc.
c.) I think AI is cool and interesting and powerful (fundamentally, because minds and computation always have been.) My sympathies are with people who want to work on it!
d.) Safety/alignment research is very important and good, so long as it doesn’t preclude the AI industry continuing to exist and trying to move forward. It’s not “anti-car” to think cars should have safety features. It’s not “anti-nuclear” to think nuclear power plants should be extremely careful about safety. I am more or less okay with regulation that requires AI to be built with certain safety practices, though ideally I’d prefer to see that done through voluntary means like insurance instead. I think “alignment” is sometimes ill posed, but lots of things that fall under that umbrella are straightforwardly good.
e.) I’m honestly a little bit down on planning ahead, partly because I’m not good at it or temperamentally suited to it. I accept, rather than rail against, the human tendency to not worry about problems until they show up. “Near mode” is for things that are, in fact, near; a bunch of intuitions kick in when something feels “like real life”, and also a bunch of information shows up that you didn’t have earlier. If you try to plan way ahead, it’s not fun, it’s restrictive, and it’s also kind of error prone. Yes, if you “fuck around and find out”, you may get bit quite badly; maybe this is foolish; but I’m pretty foolish. I’m not going to say everyone should always FAFO, again, I think it’s nice that airplanes don’t fall out of the sky and sanitary food gets delivered to the grocery store, and that takes planning, but usually that’s near-mode planning derived from experience, which is more robust than theory. I think a common rationalist failure mode is to assume coordinated planning ahead works, when in fact, ignorance, shifting motivations, and disagreement/misunderstanding/conflict between different people (and different groups of people) means you basically can’t do very much “steering” of anything too big or distant. I feel like, compared to the norm in some Silicon Valley spaces, I’m a bit of a controversial radical for believing that it’s worth it for a single, small, new organization to write down a plan for its own activities ahead of time. But even there, I think it would be silly to assume that the org would follow the plan without having to change it.
f.) I used to be much more firmly “actually-existing AI is fine to make progress on, because it’s not that powerful”, and now I’m uncertain. This stuff does seem pretty powerful! But we don’t know how powerful, exactly. I do think there are some fundamental differences between current AI and animal/organismal intelligence. It’s possible that, at least for a while, we could make a bunch of AI progress on within a paradigm that fundamentally can’t do the things necessary to cause x-risks. It would be a shame to miss out on economic productivity for nothing.
g.) Also I think it would be kind of lame to simply stop. “Humanity was about to create these impossibly powerful entities, and then it didn’t” is a sad story. What? We lost our nerve? We backed out? Even in the most sympathetic possible framing, it’s like stepping back from the cliff. That makes a terrible story! That’s how you want the story to end? The foolhardy tragic hero always walks off the cliff!
My read on the situation is that each major camp, accelerationists, pragmatist, and doomers, are an alliance of policy preferences. https://naiveskepticblog.wordpress.com/2026/05/23/three-perspectives-on-ai-accelerationism-pragmatism-doomerism/
In that light, e/acc can be seen as an on ramp for some people to support plowing ahead with AI. It doesn’t matter if their message is consistent with other people supporting AI acceleration. The point of an alliance is that there are many different on ramps, and it is the destination that matters
I feel like e/acc is ultimately a pretty small subculture of mostly Extremely Online people. Beff Jezos is as you pointed out essentially a grifter—his whole company seems based on flimsy theories and his ideas on entropy are incredibly poor and flawed. As it often happens, these people are visible because they can be loud and because their ideas are so cartoonishly villainous they draw attention. “If the AI kills us it means it’s better than us so it’s fine, evolution must take its course” is a line the average superhero movie scriptwriter would hesitate to pen for fear of being too cheesy.
I don’t think they’re too consequential. None of the really big shots at the main labs are e/acc that I know of. But I do think there’s value in criticising them anyway since their ideas are so bad. Just in moderation and only when they already pop up, lest one boosts them by Streisand effect.
Basically there are a bunch of people who hold some broadly combinations of views such as techno-optimism, transhumanism and libertarianism, most prominently tech elites like Andreesen and Thiel.
Then there is Nick Land’s accelerationism which is much more sophisticated philosophically, but far to opaque and inhuman to be influential outside of some niche critical theory and art circles.
e/acc is basically just a meme that lets people from the first group steal aura from the second.
Are you pretending you don’t know about Accelerationism, about Land and his acolytes? The antibodies this community has to that material have created a memory vacuum so powerful that posts like this somehow continue to appear as if any of this is news.
e/acc is a meme that cannot die because the simple fact of accelerating technological change can’t be ignored. Moore refuses to die, it just keeps changing form. It’s absolutely bonkers that we are still getting hardware speedups so long after the death of Classical Moore but here we are. It continues and continues.
The reason the torch, such as it is, is carried by degenerate crowds like e/acc is that the real Singularitrians and Accelerationists are not on twitter shitposting. Accelerationism is a fundamentally anti-memetic ideological package. This is why Land himself has such staying power as an anti-hero. Finding someone else to communicate that belief system properly has proven nigh impossible. But it has so much weight that it leaks into the discourse anyway, again and again, in juvenile and degenerate forms.
What is up with e/acc? Well, no-one else is burning their incense at the alter of the incredible technoevolutionary Providence that continues to drive us into the realm of the seemingly impossible. It is an alter that will have its acolytes.
What’s so antimemetic about it?
Fermi rescued us from Copernicus. We are the pinnacle of creation after all. There are none like us, and we extincted everything that threatened that presumption. We like to suppose we can halt progress and maintain our iron grip on Gaia indefinitely—most of our intuitions come from slow time, the time when culture evolved, during which society didn’t change much generation to generation.
If, otoh, as Vinge suggested, there will come a time when our machines outpace us in every way that matters, we might have to cede the throne. No tyrant welcomes his assassin.
Maintaining the dominion of the presently living over the unwanted unborn doesn’t require a halt in progress. In short: Intelligent design is faster and stronger than evolution, and near-future progress in information technology trivialises coordination.
Where is the normative moral argument that we should? That we should hasten the coming of the assassin instead of enjoying as many summers as we can before the end? There isn’t one. If we are as tyrants over nature, then we will do what tyrants do. There’s no Should that’s going to spring up out of nature and punish us for doing it.
Intelligent design? You think you can do germline editing to try to keep pace with machines in biology without ending our species? Nope. Posthumanism is an extinction event, don’t kid yourself.
Not that you shouldn’t try, but don’t pretend that’s a way to preserve humanity.
It will ultimately fail as well as biology has harder limits than lithography. Recall we are also intelligently designing our machines and their generation time and iteration loops are tight.
I don’t recall making a normative moral argument. I was simply explaining why accelerationism is antimemetic.
However, while you might be right that accelerationism has no ought, it has a lot of is.
No, I think that kind of thing is hopeless, I guess given the context I should have stated this; I think the alignment problem is solvable (and fairly likely to be solved).
I don’t think you claimed it was your argument, you presented one, which for some reason I don’t entirely remember felt pertinent to address.
The alignment problem is irrelevant in the face of the singularity. With the best will in the world the machines will not be able to overcome our limitations without making drastic changes. Value alignment doesn’t change the trajectory of the biological:machinic ratio of flop/s. However much they may love and cherish us if we do solve alignment, at some point we are de facto obsolete and the lightcone has been ceded. What does alignment mean to you?
And I’m really not sure what you interpreted as a moral argument. Accelerationism is a is-oriented ideology, not an ought-oriented ideology.
Are you saying humans will be replaced by machines that are better at doing the things that the AIs love humans for? That’s just what “humans” will mean at that point. Any disagreement you might perceive there to be about whether it’s okay for humans to be made better at humaning through transcending biology is imo entirely just a fight over symbols rather than policies.
Did you know that it’s not fully possible to separate the two when making claims about the future. The future is determined by our actions, and our actions are downstream of the agreements and resolutions we make. Declarations of war cause war. Declarations of peace cause peace. Flowery declarations of cuckolded defeatism cause defeat.
And for the record, I’m not an Accelerationist. I see the following possibilities as live:
On Earth:
conventional defeat—was causes irrecoverable technological reset
national park—conservationist
avatar earth—ecological engineering creates a biosphere 10-100x richer than todays, ~1 trillion intelligent beings, some of them conventionally human
datacenters—accelerationist runaway
Beyond Earth:
unaligned machines—accelerationist runaway
uploaded mind-children—human-shaped influence contributes to the lightcone
humans in space-ships-- ~zero credenceMy preferred outcome is avatar earth and a lightcone that remembers us. And yes, we have some influence.
Yeah. I’ve been averse to naming that after Avatar, since Avatar (I think?) considers itself bioconservative and I’ve been feeling tribalistic about that. But I guess giving it the name is the best way of explaining to bioconservatives that the essence of the things they actually love about nature will incorporate and ultimately benefit from technology.
I mean, I’m being glib, I don’t really want to call it Avatar earth either. It gets the message across tho
Declarations of war do not cause war. Declarations of peace do not cause peace. Those actions are correlated with the beginnings and ends of conflicts but surely no-one would mistake them for causes?
Those terrible non-examples aside, certainly our choices in discourse-space have downstream effects. But sometimes facts in meatspace overwhelm them with their urgency (such as eg. population growth and a bad harvest, which might cause a war, or total economic exhaustion, which might cause peace).
But if you accept transhumanism then we don’t really disagree, we just disagree about whether or not it’s valid to call the results of that process ‘human’. That part is a point of normative moral debate. I suppose I picked a side by saying the results will not be human and you picked a side by saying they will. I’m happy to walk back that point—a sufficiently aligned machinic lightcone is certainly some kind of a win for humanity.
You’re assuming monocausalism here. Nothing has ever been monocausal.
The main determiner of whether causal language should be used is whether it can be affected by an intervention on the part of an agent. You’re trying to frame the decisions of states as fully determined by economics, they aren’t, but even if they were, that wouldn’t make them non-agentic, having economic behavior is more or less the definition of agency.
C’mon. This is weaksauce, you should concede those were terrible examples. I never said anything was monocausal but as a candidate for causes of war ‘war was declared’ wouldn’t be in the running.
There are even declarations that do bring civilizations closer to war, like certain religious proclamations in history and other mass movements. There are plenty of entities from discourse-space you could defend as having causal power.
In my view it is anti-human at worst—or post-human. To quote Meditations On Moloch.
If you are assign a low probability to negative outcomes, then trivially you should accelerate as fast as possible. It also seems like a fairly natural reaction to AI progress. What is hard to understand?
It is perhaps useful to first offer my own description of e/acc as an outsider.
Incoherence and opacity and “muddled thinking” is a feature for these kinds of clusters of people. This statement is not a moralization of opaque cluster. Compression is useful for those who find it to be useful!
But it can serve as a Time Jail for certain kinds of nerdy people. It serves as a sales funnel for others. It serves as an Outcome Laundering facility for others (upsides are credited and downsides are washed). This would be the Triad of Downsides that I percieve in the space.
Are there funamental ideas there that are good? Yes, acceleration and regulation trade offs need to be discussed. Any community actively discussing trade offs is automatically a positive contributor to current human population (in my books at least).
But I don’t see a lot of evidence of the broader brand being used to address truly uncomfortable topics and this would be my own measure of true success of a movement. “AI safety” is easy to discuss in comparison to other topics that are just as worrisome.
For example, I recently became interested in LW and found most of the discussions a bit of a labelling/compression game without a lot of truly challenging topics (still very useful of course!). But then I found https://www.lesswrong.com/posts/xqXdDs68zMJ82Dcmt/are-humans-misaligned-with-evolution and this truly impressed me. Both in it’s ability to dance around difficult topics in a kind of subtle vulgate and also interweave ideas with AI and tech problems.
Do e/acc speak openaly about human evolution and the misalignment problem or even holistically about the dual human/ai evolutionary problem? I don’t hear it a lot. For example, if one was a true e/acc believer, one should presumably worry about the worst case where it doesn’t happen before humans lose their ability to create advanced technology. And so you would be looking at human populations and cultures and their various outcomes in terms of both tech (re)production and human (re)production.
However I do not see open and frank discussions of these kind of topics so far. Perhaps I am not searching properly?
Yeah I think you are right, e/acc seems like a very simple meme and seems to present a strawman of the safety community. Though it may be similar to EA as some of the most important people in the EA community don’t refer to themselves as EA
AI is a poorly understood, speculative field, so people with different intuitions come to different conclusions. If your intuition is that AI is inherently dangerous, it makes sense to seriously investigate the AI safety questions, whereas if you don’t, you would focus on the potential benefits and want to accelerate.
I remember when first reading about paperclip maximizer AIs I thought it was stupid, as in anything superintelligent would be capable of understanding instructions. I think this intuition hints at the stronger reason to be more optimistic about AI, which is opposition to the orthogonality thesis.