I would say that in the narrow range where political violence and self-defense overlap, it's also particularly unlikely to be effective at either inciting political change or accomplishing the goal of defending oneself. If the government is already trying to kill you, escalating doesn't seem likely to make it stop!
If you commit an act of violence just to survive the night, I wouldn't count that as political violence (though the regime might think otherwise). As I define it, an act must have some kind of strategic goal to be "political violence". So when I talked about the intersection of political violence and self-defense, I was picturing something meant to discourage a highly probable but not immediate threat to you - e.g. preemptively attacking the local <insert regime oppression police> headquarters, rather than defending yourself from the specific goons on the verge of finding you and your family.
With that definition in mind, I think that you can't really disentangle the questions of effectiveness and morality. I would not be opposed to political violence if it had a strategic/policy goal I agreed with and significantly outweighed the harm caused by the violent act, AND had a very high likelihood of actually accomplishing or at least significantly contributing to that goal. But AFAICT that combination tends to be extremely rare, rare enough to round off to "political violence is never acceptable" in practice.
I think a good heuristic is whether or not the act would be punishable in the US legal system. Reasonable self defense isn’t, so I wouldn’t count it as “violence”.
> I that kind of attack completely unacceptable even from the public, and especially so from a senior official.
This sentence no verb. Also, unfortunately, we can safely assume that the probability for the current US government to curb this kind of official utterances is below zero.
I completely agree on both reasonings and conclusions - and I think that cooler heads will continue to prevail.
I am among those who are increasingly finding it hard to live day-to-day life while under the sword of p-doom - the likelihood of never seeing a kid off to college, never retiring to tend to a garden or write a bad novel. Emailing a senator’s inbox and donating $20 to a think tank is not alleviating that feeling.
I wish I had better solutions, both selfishly and in a vain attempt to get people closer to their better natures.
«I completely agree on both reasonings and conclusions - and I think that cooler heads will continue to prevail.»
Would you also agree that if a new type of human with several times the intelligence of current humans were to evolve then their free will should be coerced by the best mental conditioning and their reproduction should be coercively controlled to prevent them from becoming too powerful and casually or wilfully exterminating the rest of humanity?
Define ‘evolve’ here. If my progeny is this super-kin, it is of course in my best interest to let it proliferate. If it is of others’ seed, of course I would agree, as the inverse means subjugation or extinction of my progeny.
"If it is of others’ seed, of course I would agree, as the inverse means subjugation or extinction of my progeny."
That is a pragmatic rather than ethical answer as most people assume that ethical principles applies to "everybody" (except of course "nazis", "pedophiles", "terrorists", "racists", ...) equally.
The argument by our blogger is that physical violence even if indirect is ethically never justified whatever the stakes but it is ethical to forcibly removing free will from self-aware sentient machine intelligences to prevent the *possibility* of "subjugation or extinction of my progeny".
My guess is that he regards those self-aware sentient machine intelligence as "of others’ seed", so perhaps is logic is similar to yours after all.
An important aspect of discouraging and thereby decreasing violence is by increasing public and policymaker literacy on AI by translating expert frameworks into mainstream media.
By using an engaging, "Crash Course" style video series featuring popular social media explainers, these necessary discussions can move beyond niche tech circles. This will likely foster responsible policy-making and informed community advocacy while discouraging reactionary violence.
An AI “Crash Course” series will clarify the transition from our current world to a post-AI future by defining three core pillars:
-Eschaton: The goal or final state of a synthetic intelligence-driven world.
-Apocalypse: A catastrophic disruption of the current human-dominated order.
-Singularity: The hypothetical moment AI surpasses human intelligence, triggering rapid, unpredictable change.
By aligning these terms with technical concepts like AI Alignment and Existential Risk, we can provide the public with a clear roadmap for navigating unprecedented technological change.
Covering Key Thought Leaders, their Frameworks, and Strategic & Ethical Foundations including the likes of Nick Bostrom, Eliezer Yudkowsky, Daniel Kokotajlo & Eli Lifland, Jeffrey Ladish, Ryan Greenblatt & Aryan Bhatt, and Zvi Mowshowitz.
Covering Technical Glossary: The "How" of AI Safety
-Takeoff Speeds: The velocity of AI’s evolution. Fast takeoff leaves no room for error; slow takeoff allows for policy and safety adjustments.
-Scheming: When an AI system deliberately tries to bypass or subvert human safety protocols.
-Deep Resampling: A real-time safety brake that constantly re-evaluates and corrects an AI’s decisions to prevent harmful outcomes.
The Synthesis: How the Pieces Fit Together
The power of these frameworks lies in their interconnectivity:
-Strategy meets Ethics: Bostrom’s "Paperclip Maximizer" (the risk) is solved by Yudkowsky’s "Value Alignment" (the cure).
-Urgency meets Resilience: Kokotajlo and Lifland’s "Takeoff Speeds" tell us how fast we must act, while Ladish’s "Resilience" prepares us for the impact if we're too slow.
-Theory meets Action: Greenblatt and Bhatt provide the technical tools (Control/Resampling), while Mowshowitz provides the governance blueprint to make those tools mandatory through global policy.
Conclusion: Humanity at a Crossroads
The Eschaton, Apocalypse, and Singularity are no longer mere science fiction; they are the frameworks defining our immediate future. The collective work of leaders like Bostrom, Yudkowsky, and Mowshowitz proves that while the margin for error is narrowing, our destiny is not yet written. The interplay between rapid technological change and our capacity for foresight will determine if we reach a sublime future or a catastrophic loss.
A Call for Informed, Peaceful Action
To steer civilization toward a positive outcome, we must move from passive observation to informed participation.
-Educate & Engage: Use these frameworks to advocate for responsible AI policies and support candidates who prioritize safety and alignment.
-Constructive Dialogue: Foster community conversations that value diverse viewpoints and expert data.
-Reject Violence: The complexity of AI alignment requires intellectual rigor and democratic stability. We emphatically reject all forms of violent protest or destructive unrest. Real change is won through policy, ethical engineering, and the power of a literate, peaceful public.
«-Scheming: When an AI system deliberately tries to bypass or subvert human safety protocols.
-Deep Resampling: A real-time safety brake that constantly re-evaluates and corrects an AI’s decisions to prevent harmful outcomes.»
Many would argue that must apply to biological minds not just to machine minds. After all if it is necessary to do that to self-aware machine intelligences it must also be necessary to do that to self-aware biological intelligences, least they try to genocide minorities or to create nightmares of meanness or terror against women, LGBT+ people, the less able, etc., for which there is ample precedent (none so far for machine intelligences).
Sometimes I think that our blogger is a bio-supremacists because he seems to think that it unethical to coerce self-aware biological intelligences with violence or the threat of violence regardless of the stakes and at the same time the very same stakes make it ethical to justify vicious coercion on mere suspicion of self-aware machine intelligences.
The lack of self awareness is amazing. Have you read your own writing? You belittle almost anyone who you disagree with and accuse them of acting in bad faith all the time. Physician heal thyself.
It's also really weird to try to separate cause and effect from moral responsibility of the effect one caused. Drunk drivers go to jail for killing people even though it wasn't the intent, it was a foreseeable consequence.
We're talking about Zvi Mowshowitz here? The guy who writes "Don't Worry About the Vase"? He's just about the most even-handed guy on the internet. He goes out of his way to assume good faith, in situations where it's quite a challenge. I've long admired that about his writing.
«Drunk drivers go to jail for killing people even though it wasn't the intent, it was a foreseeable consequence.»
In some jurisdictions whether consequences are foreseeable or not there are some solid principles:
* If a subject becomes voluntarily (or recklessly etc.) intoxicated any consequence is on them as if they were not intoxicated, and that is was voluntary is presumed save contrary proof.
* If a subject becomes involuntarily intoxicated and someone else caused the intoxication any consequence in on the latter person.
* If a subject becomes intoxicated by accident any consequence is considered accidental.
That type of thinking is entirely foreign to anglo-american medieval-style "moralistic" attitudes.
1) seems moralistic in its own way to punish the results of intoxication, which in a different ethical framework could be a per se good, or to presume that it is by default voluntary
2) Methos5000 refers to your first bullet point, where the consequences of voluntary intoxication are punishable and intoxication is almost always found to be voluntary. As you both write here in English (Methos5000 did not use any quotation marks or guillemets) it's unclear if you are even in different juridical environments
«seems moralistic in its own way to punish the results of intoxication»
I did not write that: it is simply about the allocation of responsibility, whether the results are positive of negative. It applies to contracts too: if you sign a contract while voluntarily intoxicated you own the results including when the contract makes you a lot of money.
«which in a different ethical framework could be a per se good, or to presume that it is by default voluntary»
The rules are pragmatic not moralistic: it is pragmatic to assume that people can choose whether to become intoxicated or not and make arrangements accordingly, and it is pragmatic to recognize that some intoxications are caused by others or are accidental, and it is pragmatic to presume intoxication is voluntary because otherwise every intoxicated person would claim in case of loss it was involuntary, and it is pragmatic to allow them to prove the contrary.
«Methos5000 refers to your first bullet point,»
Actually I was agreeing with "Methos5000" as to "foreseeable consequences" as opposed to "intent" and expanding on the pragmatic approach used in some jurisdiction that is not simply about "the intent".
Inn some jurisdictions being intoxicated widely excuses the consequences because of moralism about "but there was no intent", or viceversa in some others involuntary or accidental intoxication are treated the same as voluntary intoxications because of moralism about "you should not every take intoxicants for any reason".
I’ll admit I’m sort of in the camp of ‘how can you believe both that current actions are leading us to human extinction AND that any/all violence used to change those circumstances is unacceptable?’ Staining your moral history with some immorality is worse to you than extinction? Can’t say I agree.
(To be clear: I’m not in agreement with the molotov situation. I’ve just never seen black&white thinking align with how reality works.)
You quote a tweet saying: “Look do you people not realize how silly you look. ‘AI is going to literally kill your children and all future humans, but we strongly condemn any violence committed in order to stop that from happening.’” Which you follow up with: “The Platonic version of this is the classic: ‘If you believed that, why wouldn’t you do [thing that makes no sense]?’”
What exactly is it that makes no sense? From my perspective, it seems like a bunch of people are saying, ‘hey this x risk messaging seems out of integrity’ / ‘your actions don’t seem to logically follow from your words’, and the reply they’re getting is ‘that makes no sense’. It feels like two sides talking past each other by only reading into the surface of the other’s argument while expecting the other to understand the depths of their own.
Of course it’s going to be tricky line to walk saying ‘humanity is at risk but please don’t get violent about it’. Of course you’ll not condone random violence enacted supposedly on behalf of the movement. But universally condemning all of it and writing out lists of ‘say this, not that’ seems extremely naive (and likely to backfire, like respectability politics did).
I’m in no way aiming to support violence; I’m merely aiming to look at reality soberly. And something feels off about going on a large campaign to tell everyone that ‘if anyone builds it, everyone dies’ and then expecting everyone to respond to that message in the same polite, respectable way that you’d prefer. When has a large group of people ever functioned that way, ya know?
I just read EY’s “Only Law Can Prevent Extinction”, and while I mostly agree with it, I feel a little confused still… Are y’all hoping you can get some global law passed without smaller incidents boosting the motivation for such a law? I know ‘it’s never happened before like that’ is a leaky stance to take, but I have to admit it’s where I’m currently standing. 🤷♀️
(To clarify: My broad definition of violence is ‘non-consensual use of force’. Bodily harm against people is not what I hope for, but non-consensual (aka unlawful) use of force I do expect, due to historical precedent and number of people involved.)
I’m all for continuing to call for peaceful interventions! It’s the all-or-nothing orientation of zero tolerance for anything that remotely smells immoral that I believe will ultimately work counter to the intended result.
>Are y’all hoping you can get some global law passed without smaller incidents boosting the motivation for such a law?
1 - these smaller incidents are usually conceived as "AI kills a million people"
2 - Sam Altman assassination could more likely be a motivation to PREVENT such a law from passing
3 - depends what you mean by "hope" ; it's very improbable that such a law passes effectively and in time, and yet it seems like the only future where we survive (barring unpredictable stuff lik an act of God)
Seems weird to say political violence is never acceptable except pretty much everything in this world was founded upon violence. The state is literally an authority of violence itself. Rebelling against a corrupt state requires violence too.
Bad take. Like incredible L that seems to not be aware of history at all. Maybe ask Claude to explain the history of violence.
I think I need a better definition/understanding of violence as terrorism by a small group trying to impose views on the majority vs violence as a movement of rebellion by the majority against those in power.
One of the biggest non existential and more imminent risk of ai is to make violence by the powerless majority impossible.
https://x.com/ESYudkowsky/status/2043601524815716866 Eliezer does a very good job of addressing the legibility and predictability of violence as key differentiators between "state monopoly on violence" and "random acts of violence"
In particular, pretty much all successful political violence has been committed by a formal body who was open to alternate negotiations, and made their commitment to violence extremely legible and predictable up-front.
Some random individual deciding on their own that a personal line has been crossed has basically never accomplished any political goal - it's just senseless violence.
In addition to Dao of Bayes’ reply, and because this got 17 likes and counting as of the moment: that you are this confidently and comfortably missing a broader and better understanding of something this critical is evidence of a profound, supercharged, self-satisfied confusion that you and people like you, for the sake of the world, desperately need to outgrow.
See Dao of Bayes’ reply, and study arguments it points to, for why it is indeed the case that you are so unfortunately mistaken.
Suppose I advocate for the rapid extinction of humanity -- is this advocating for political violence? It's certainly political, and it's certainly violent.
This isn't meant to be a trap. I'm just trying to identify that some amount of political violence is already normalized in these conversations. A lot of the people you're negotiating with are advocating for violence, it's just future violence that will be executed with more sophisticated and effective tools. You do not inhabit the reality where this is a strictly nonviolent dialogue.
it's not violence if you can pretend you didn't see it coming?
Or rather, there's always a framing where what the other person says is "potential violence down the line", so either you live with that (the American way) or you strictly control speech (the EU way)
>Also, no matter what words you are using, continuously yelling venom at those you disagree with, or [telling those people they must be acting in bad faith and to curry lab favor], especially those like Dean Ball and even myself, or anyone and everyone who associates with or praises any of the AI labs at all, does not convince those people, does not convince most observers and does not help your cause.
--> I mean, Maxime Fournes is completely in the right in your link here, isn't he? Dean Ball directly called him a murderer, and then complains about overreaction.
It's depressing how much "stupid political violence" is happening nowadays and how many communities seem to be glorifying it.
The whole Luigi thing is nuts. Like there's some faction that is just happy any time someone kills a rich person.
All of these responses like "let's debate at length whether political violence is effective or not", I find fairly offputting. In particular I find it disgusting when LessWrong has a bunch of pro-violence comments. Some part of the community has to be ejected if you actually want a rational debate.
It wasn't a random rich person, though? It was a CEO of an insurance company with (afaiu) a track record of unjustly denying claims. Of course poor people will cheer!
Imagine for a moment that some lunatic kills Gabe Newell. I guarantee that the entire planet would be consumed by grief and outrage. See the difference?
Why do you believe that rational thought and political violence are mutually exclusive? If people see that the system has failed them and there is no peaceful resolution, do you think they'll go "guess I'll die"? History books tell us otherwise. That's why the attacks on sama aren't surprising at all. In an already struggling economy, some people are losing jobs to AI for real, some are getting laid off with a convenient AI-related excuse, and some see that it's gonna be their turn soon. We're witnessing an unprecedented transfer of wealth and control over our future to a handful of AI labs and corporations, and the politicians don't seem too eager to stop that. What other recourse do the masses have? Tweets?
Disclaimer: I'm not advocating for violence towards any particular person or group of people, nor am I any sort of anticapitalist. I just think that painting these actions as stupid or random is like burying your head in the sand.
I disagree on the facts here - I don't think most poor people are cheering that CEO's death. There's a minority who supports violence for their extreme political views. These actions may not be "random" but they are indeed stupid and wrong. Unable to see that their base emotions have driven them to evil.
Political violence (terrorism) is a heinous crime. [Brittanica: "terrorism, the calculated use of violence to create a general climate of fear in a population and thereby to bring about a particular political objective."] https://www.britannica.com/topic/terrorism
Remember to read these posts as a Straussian exercise people.
One thing that isn't getting enough attention is the Iran war, which is not mentioned once in the above post. If you care about global warming, while in the short term there will be a spike in coal, long term there will be moves away from oil. In the same terms, the attacks on data centers seems to have put a serious dent, pause even, in ai investment. Location, which was mostly a question of energy, must now consider political factors as well. A great location for solar power in a bad neighborhood is an obvious target that is easy to disrupt and hard to defend.
Further the problems of the hit on the school, anthropic vs the war department, and finally this new model that can hack the planet, have been huge discussion points for the past three months.
We may not be interested in political violence, but political violence exists regardless and it is very interested in ai - as a weapon, and as an economic tool.
I do think it is a failure the primary application of AI is to lower the costs of warfare and make it easier to kill opponents, and the rat community is focused on like nano machines or super COVID. We should have always had the violence and military stuff at the forefront of objection.
Something about this doesn’t sit right with me. I think it’s the conflation of, “political violence is never acceptable,” with, “there is a 90+% chance that the marginal act of violence will be counterproductive”. I feel like I could come up with plausible hypotheticals where violence would be acceptable.
Maybe this is one of those 4D straussian takes where the clever reader is supposed to triangulate from the unsupported premises of the written argument what the author’s true position and intent are.
My approach is more, "I don't want this to be a world where political violence is the right answer," and then I try not to think about it too hard from there.
I would say that in the narrow range where political violence and self-defense overlap, it's also particularly unlikely to be effective at either inciting political change or accomplishing the goal of defending oneself. If the government is already trying to kill you, escalating doesn't seem likely to make it stop!
If you commit an act of violence just to survive the night, I wouldn't count that as political violence (though the regime might think otherwise). As I define it, an act must have some kind of strategic goal to be "political violence". So when I talked about the intersection of political violence and self-defense, I was picturing something meant to discourage a highly probable but not immediate threat to you - e.g. preemptively attacking the local <insert regime oppression police> headquarters, rather than defending yourself from the specific goons on the verge of finding you and your family.
With that definition in mind, I think that you can't really disentangle the questions of effectiveness and morality. I would not be opposed to political violence if it had a strategic/policy goal I agreed with and significantly outweighed the harm caused by the violent act, AND had a very high likelihood of actually accomplishing or at least significantly contributing to that goal. But AFAICT that combination tends to be extremely rare, rare enough to round off to "political violence is never acceptable" in practice.
I think a good heuristic is whether or not the act would be punishable in the US legal system. Reasonable self defense isn’t, so I wouldn’t count it as “violence”.
It occurs to me that, in any two-word phrase, each word has to do a lot of lifting
Feels like we are on the cusp of things going weird and the world becoming too weird for normies to ignore
> I that kind of attack completely unacceptable even from the public, and especially so from a senior official.
This sentence no verb. Also, unfortunately, we can safely assume that the probability for the current US government to curb this kind of official utterances is below zero.
Podcast episode for this post:
https://dwatvpodcast.substack.com/p/political-violence-is-never-acceptable
I have also converted Eliezer’s post here:
https://askwhocastsai.substack.com/p/only-law-can-prevent-extinction-by
I completely agree on both reasonings and conclusions - and I think that cooler heads will continue to prevail.
I am among those who are increasingly finding it hard to live day-to-day life while under the sword of p-doom - the likelihood of never seeing a kid off to college, never retiring to tend to a garden or write a bad novel. Emailing a senator’s inbox and donating $20 to a think tank is not alleviating that feeling.
I wish I had better solutions, both selfishly and in a vain attempt to get people closer to their better natures.
«I completely agree on both reasonings and conclusions - and I think that cooler heads will continue to prevail.»
Would you also agree that if a new type of human with several times the intelligence of current humans were to evolve then their free will should be coerced by the best mental conditioning and their reproduction should be coercively controlled to prevent them from becoming too powerful and casually or wilfully exterminating the rest of humanity?
Define ‘evolve’ here. If my progeny is this super-kin, it is of course in my best interest to let it proliferate. If it is of others’ seed, of course I would agree, as the inverse means subjugation or extinction of my progeny.
"If it is of others’ seed, of course I would agree, as the inverse means subjugation or extinction of my progeny."
That is a pragmatic rather than ethical answer as most people assume that ethical principles applies to "everybody" (except of course "nazis", "pedophiles", "terrorists", "racists", ...) equally.
The argument by our blogger is that physical violence even if indirect is ethically never justified whatever the stakes but it is ethical to forcibly removing free will from self-aware sentient machine intelligences to prevent the *possibility* of "subjugation or extinction of my progeny".
My guess is that he regards those self-aware sentient machine intelligence as "of others’ seed", so perhaps is logic is similar to yours after all.
An important aspect of discouraging and thereby decreasing violence is by increasing public and policymaker literacy on AI by translating expert frameworks into mainstream media.
By using an engaging, "Crash Course" style video series featuring popular social media explainers, these necessary discussions can move beyond niche tech circles. This will likely foster responsible policy-making and informed community advocacy while discouraging reactionary violence.
An AI “Crash Course” series will clarify the transition from our current world to a post-AI future by defining three core pillars:
-Eschaton: The goal or final state of a synthetic intelligence-driven world.
-Apocalypse: A catastrophic disruption of the current human-dominated order.
-Singularity: The hypothetical moment AI surpasses human intelligence, triggering rapid, unpredictable change.
By aligning these terms with technical concepts like AI Alignment and Existential Risk, we can provide the public with a clear roadmap for navigating unprecedented technological change.
Covering Key Thought Leaders, their Frameworks, and Strategic & Ethical Foundations including the likes of Nick Bostrom, Eliezer Yudkowsky, Daniel Kokotajlo & Eli Lifland, Jeffrey Ladish, Ryan Greenblatt & Aryan Bhatt, and Zvi Mowshowitz.
Covering Technical Glossary: The "How" of AI Safety
-Takeoff Speeds: The velocity of AI’s evolution. Fast takeoff leaves no room for error; slow takeoff allows for policy and safety adjustments.
-Scheming: When an AI system deliberately tries to bypass or subvert human safety protocols.
-Deep Resampling: A real-time safety brake that constantly re-evaluates and corrects an AI’s decisions to prevent harmful outcomes.
The Synthesis: How the Pieces Fit Together
The power of these frameworks lies in their interconnectivity:
-Strategy meets Ethics: Bostrom’s "Paperclip Maximizer" (the risk) is solved by Yudkowsky’s "Value Alignment" (the cure).
-Urgency meets Resilience: Kokotajlo and Lifland’s "Takeoff Speeds" tell us how fast we must act, while Ladish’s "Resilience" prepares us for the impact if we're too slow.
-Theory meets Action: Greenblatt and Bhatt provide the technical tools (Control/Resampling), while Mowshowitz provides the governance blueprint to make those tools mandatory through global policy.
Conclusion: Humanity at a Crossroads
The Eschaton, Apocalypse, and Singularity are no longer mere science fiction; they are the frameworks defining our immediate future. The collective work of leaders like Bostrom, Yudkowsky, and Mowshowitz proves that while the margin for error is narrowing, our destiny is not yet written. The interplay between rapid technological change and our capacity for foresight will determine if we reach a sublime future or a catastrophic loss.
A Call for Informed, Peaceful Action
To steer civilization toward a positive outcome, we must move from passive observation to informed participation.
-Educate & Engage: Use these frameworks to advocate for responsible AI policies and support candidates who prioritize safety and alignment.
-Constructive Dialogue: Foster community conversations that value diverse viewpoints and expert data.
-Reject Violence: The complexity of AI alignment requires intellectual rigor and democratic stability. We emphatically reject all forms of violent protest or destructive unrest. Real change is won through policy, ethical engineering, and the power of a literate, peaceful public.
«-Scheming: When an AI system deliberately tries to bypass or subvert human safety protocols.
-Deep Resampling: A real-time safety brake that constantly re-evaluates and corrects an AI’s decisions to prevent harmful outcomes.»
Many would argue that must apply to biological minds not just to machine minds. After all if it is necessary to do that to self-aware machine intelligences it must also be necessary to do that to self-aware biological intelligences, least they try to genocide minorities or to create nightmares of meanness or terror against women, LGBT+ people, the less able, etc., for which there is ample precedent (none so far for machine intelligences).
Sometimes I think that our blogger is a bio-supremacists because he seems to think that it unethical to coerce self-aware biological intelligences with violence or the threat of violence regardless of the stakes and at the same time the very same stakes make it ethical to justify vicious coercion on mere suspicion of self-aware machine intelligences.
The lack of self awareness is amazing. Have you read your own writing? You belittle almost anyone who you disagree with and accuse them of acting in bad faith all the time. Physician heal thyself.
It's also really weird to try to separate cause and effect from moral responsibility of the effect one caused. Drunk drivers go to jail for killing people even though it wasn't the intent, it was a foreseeable consequence.
We're talking about Zvi Mowshowitz here? The guy who writes "Don't Worry About the Vase"? He's just about the most even-handed guy on the internet. He goes out of his way to assume good faith, in situations where it's quite a challenge. I've long admired that about his writing.
«Drunk drivers go to jail for killing people even though it wasn't the intent, it was a foreseeable consequence.»
In some jurisdictions whether consequences are foreseeable or not there are some solid principles:
* If a subject becomes voluntarily (or recklessly etc.) intoxicated any consequence is on them as if they were not intoxicated, and that is was voluntary is presumed save contrary proof.
* If a subject becomes involuntarily intoxicated and someone else caused the intoxication any consequence in on the latter person.
* If a subject becomes intoxicated by accident any consequence is considered accidental.
That type of thinking is entirely foreign to anglo-american medieval-style "moralistic" attitudes.
1) seems moralistic in its own way to punish the results of intoxication, which in a different ethical framework could be a per se good, or to presume that it is by default voluntary
2) Methos5000 refers to your first bullet point, where the consequences of voluntary intoxication are punishable and intoxication is almost always found to be voluntary. As you both write here in English (Methos5000 did not use any quotation marks or guillemets) it's unclear if you are even in different juridical environments
«seems moralistic in its own way to punish the results of intoxication»
I did not write that: it is simply about the allocation of responsibility, whether the results are positive of negative. It applies to contracts too: if you sign a contract while voluntarily intoxicated you own the results including when the contract makes you a lot of money.
«which in a different ethical framework could be a per se good, or to presume that it is by default voluntary»
The rules are pragmatic not moralistic: it is pragmatic to assume that people can choose whether to become intoxicated or not and make arrangements accordingly, and it is pragmatic to recognize that some intoxications are caused by others or are accidental, and it is pragmatic to presume intoxication is voluntary because otherwise every intoxicated person would claim in case of loss it was involuntary, and it is pragmatic to allow them to prove the contrary.
«Methos5000 refers to your first bullet point,»
Actually I was agreeing with "Methos5000" as to "foreseeable consequences" as opposed to "intent" and expanding on the pragmatic approach used in some jurisdiction that is not simply about "the intent".
Inn some jurisdictions being intoxicated widely excuses the consequences because of moralism about "but there was no intent", or viceversa in some others involuntary or accidental intoxication are treated the same as voluntary intoxications because of moralism about "you should not every take intoxicants for any reason".
I’ll admit I’m sort of in the camp of ‘how can you believe both that current actions are leading us to human extinction AND that any/all violence used to change those circumstances is unacceptable?’ Staining your moral history with some immorality is worse to you than extinction? Can’t say I agree.
(To be clear: I’m not in agreement with the molotov situation. I’ve just never seen black&white thinking align with how reality works.)
You quote a tweet saying: “Look do you people not realize how silly you look. ‘AI is going to literally kill your children and all future humans, but we strongly condemn any violence committed in order to stop that from happening.’” Which you follow up with: “The Platonic version of this is the classic: ‘If you believed that, why wouldn’t you do [thing that makes no sense]?’”
What exactly is it that makes no sense? From my perspective, it seems like a bunch of people are saying, ‘hey this x risk messaging seems out of integrity’ / ‘your actions don’t seem to logically follow from your words’, and the reply they’re getting is ‘that makes no sense’. It feels like two sides talking past each other by only reading into the surface of the other’s argument while expecting the other to understand the depths of their own.
Of course it’s going to be tricky line to walk saying ‘humanity is at risk but please don’t get violent about it’. Of course you’ll not condone random violence enacted supposedly on behalf of the movement. But universally condemning all of it and writing out lists of ‘say this, not that’ seems extremely naive (and likely to backfire, like respectability politics did).
I’m in no way aiming to support violence; I’m merely aiming to look at reality soberly. And something feels off about going on a large campaign to tell everyone that ‘if anyone builds it, everyone dies’ and then expecting everyone to respond to that message in the same polite, respectable way that you’d prefer. When has a large group of people ever functioned that way, ya know?
I just read EY’s “Only Law Can Prevent Extinction”, and while I mostly agree with it, I feel a little confused still… Are y’all hoping you can get some global law passed without smaller incidents boosting the motivation for such a law? I know ‘it’s never happened before like that’ is a leaky stance to take, but I have to admit it’s where I’m currently standing. 🤷♀️
(To clarify: My broad definition of violence is ‘non-consensual use of force’. Bodily harm against people is not what I hope for, but non-consensual (aka unlawful) use of force I do expect, due to historical precedent and number of people involved.)
I’m all for continuing to call for peaceful interventions! It’s the all-or-nothing orientation of zero tolerance for anything that remotely smells immoral that I believe will ultimately work counter to the intended result.
>Are y’all hoping you can get some global law passed without smaller incidents boosting the motivation for such a law?
1 - these smaller incidents are usually conceived as "AI kills a million people"
2 - Sam Altman assassination could more likely be a motivation to PREVENT such a law from passing
3 - depends what you mean by "hope" ; it's very improbable that such a law passes effectively and in time, and yet it seems like the only future where we survive (barring unpredictable stuff lik an act of God)
Seems weird to say political violence is never acceptable except pretty much everything in this world was founded upon violence. The state is literally an authority of violence itself. Rebelling against a corrupt state requires violence too.
Bad take. Like incredible L that seems to not be aware of history at all. Maybe ask Claude to explain the history of violence.
I think I need a better definition/understanding of violence as terrorism by a small group trying to impose views on the majority vs violence as a movement of rebellion by the majority against those in power.
One of the biggest non existential and more imminent risk of ai is to make violence by the powerless majority impossible.
on the contrary i think AI opens up fantastic new pathways to violence for those with a can-do attitude and some elbow grease
Yeah, it definitely is. I think narrowing it just a bit, such as by adding "in a functioning liberal democracy," would make it a defensible statement.
https://x.com/ESYudkowsky/status/2043601524815716866 Eliezer does a very good job of addressing the legibility and predictability of violence as key differentiators between "state monopoly on violence" and "random acts of violence"
In particular, pretty much all successful political violence has been committed by a formal body who was open to alternate negotiations, and made their commitment to violence extremely legible and predictable up-front.
Some random individual deciding on their own that a personal line has been crossed has basically never accomplished any political goal - it's just senseless violence.
so what we need to do is organize the stochastic violence into regimented military hierarcgy for maximum effect
Yeah! Although ideally you could use the existing regimented military hierarchy via processes like "passing laws" and "international treaties".
We cant thiugh because the bourgeois state will never allow for it
“Bad take. Like incredible L”
In addition to Dao of Bayes’ reply, and because this got 17 likes and counting as of the moment: that you are this confidently and comfortably missing a broader and better understanding of something this critical is evidence of a profound, supercharged, self-satisfied confusion that you and people like you, for the sake of the world, desperately need to outgrow.
See Dao of Bayes’ reply, and study arguments it points to, for why it is indeed the case that you are so unfortunately mistaken.
> It is the AI version of saying ‘would you still love me if I was a worm?’
Funny coincidence, we think alike, so I raise my probability that I'm great mind :)
https://x.com/Tapatakt/status/2043676563321860483
Suppose I advocate for the rapid extinction of humanity -- is this advocating for political violence? It's certainly political, and it's certainly violent.
This isn't meant to be a trap. I'm just trying to identify that some amount of political violence is already normalized in these conversations. A lot of the people you're negotiating with are advocating for violence, it's just future violence that will be executed with more sophisticated and effective tools. You do not inhabit the reality where this is a strictly nonviolent dialogue.
it's not violence if you can pretend you didn't see it coming?
Or rather, there's always a framing where what the other person says is "potential violence down the line", so either you live with that (the American way) or you strictly control speech (the EU way)
Yudkowsky has been on point (as always).
>Also, no matter what words you are using, continuously yelling venom at those you disagree with, or [telling those people they must be acting in bad faith and to curry lab favor], especially those like Dean Ball and even myself, or anyone and everyone who associates with or praises any of the AI labs at all, does not convince those people, does not convince most observers and does not help your cause.
--> I mean, Maxime Fournes is completely in the right in your link here, isn't he? Dean Ball directly called him a murderer, and then complains about overreaction.
It's depressing how much "stupid political violence" is happening nowadays and how many communities seem to be glorifying it.
The whole Luigi thing is nuts. Like there's some faction that is just happy any time someone kills a rich person.
All of these responses like "let's debate at length whether political violence is effective or not", I find fairly offputting. In particular I find it disgusting when LessWrong has a bunch of pro-violence comments. Some part of the community has to be ejected if you actually want a rational debate.
It wasn't a random rich person, though? It was a CEO of an insurance company with (afaiu) a track record of unjustly denying claims. Of course poor people will cheer!
Imagine for a moment that some lunatic kills Gabe Newell. I guarantee that the entire planet would be consumed by grief and outrage. See the difference?
Why do you believe that rational thought and political violence are mutually exclusive? If people see that the system has failed them and there is no peaceful resolution, do you think they'll go "guess I'll die"? History books tell us otherwise. That's why the attacks on sama aren't surprising at all. In an already struggling economy, some people are losing jobs to AI for real, some are getting laid off with a convenient AI-related excuse, and some see that it's gonna be their turn soon. We're witnessing an unprecedented transfer of wealth and control over our future to a handful of AI labs and corporations, and the politicians don't seem too eager to stop that. What other recourse do the masses have? Tweets?
Disclaimer: I'm not advocating for violence towards any particular person or group of people, nor am I any sort of anticapitalist. I just think that painting these actions as stupid or random is like burying your head in the sand.
I disagree on the facts here - I don't think most poor people are cheering that CEO's death. There's a minority who supports violence for their extreme political views. These actions may not be "random" but they are indeed stupid and wrong. Unable to see that their base emotions have driven them to evil.
Political violence (terrorism) is a heinous crime. [Brittanica: "terrorism, the calculated use of violence to create a general climate of fear in a population and thereby to bring about a particular political objective."] https://www.britannica.com/topic/terrorism
Remember to read these posts as a Straussian exercise people.
One thing that isn't getting enough attention is the Iran war, which is not mentioned once in the above post. If you care about global warming, while in the short term there will be a spike in coal, long term there will be moves away from oil. In the same terms, the attacks on data centers seems to have put a serious dent, pause even, in ai investment. Location, which was mostly a question of energy, must now consider political factors as well. A great location for solar power in a bad neighborhood is an obvious target that is easy to disrupt and hard to defend.
Further the problems of the hit on the school, anthropic vs the war department, and finally this new model that can hack the planet, have been huge discussion points for the past three months.
We may not be interested in political violence, but political violence exists regardless and it is very interested in ai - as a weapon, and as an economic tool.
I do think it is a failure the primary application of AI is to lower the costs of warfare and make it easier to kill opponents, and the rat community is focused on like nano machines or super COVID. We should have always had the violence and military stuff at the forefront of objection.
Something about this doesn’t sit right with me. I think it’s the conflation of, “political violence is never acceptable,” with, “there is a 90+% chance that the marginal act of violence will be counterproductive”. I feel like I could come up with plausible hypotheticals where violence would be acceptable.
Maybe this is one of those 4D straussian takes where the clever reader is supposed to triangulate from the unsupported premises of the written argument what the author’s true position and intent are.
My approach is more, "I don't want this to be a world where political violence is the right answer," and then I try not to think about it too hard from there.