I'm not sure the left/right cultural gulf will last beyond Trump. I'm not even sure it's still alive now, with Trump's approval ratings in the toilet. The US political system takes time to cycle and isn't on the same schedule as the political pendulum.
I don't want to dogpile on the other comments (atheist, loved the book) but I think there's something interesting here.
Most science fiction tends to assume that religiosity will fade as humanity matures, and in a few thousand years we'll all have a good laugh at those silly ancient humans. This feels generally right to me. But it's not the only possible future, and Hyperion explores a far future in which religiosity becomes more ingrained.
I thought it was one of the more interesting aspects of the book, and contributed to the feeling of "not just another space opera". You don't have to appreciate religion to like the story.
This seems to only have a good track record in places with a democratic tradition. Some dictators have figured out you can just imprison and kill the opposition, and keep doing this until there is no more opposition.
The Khomeini government is not going to just say "oh, you're right" and change. Neither will the Kim or Putin governments. Sometimes - sadly - violence is the least worst answer.
Don't say sadly. Don't further the indoctrination that violence is bad.
It is a tool, it cannot be good or bad. States are the most prolific users of violence (even more when you also count potential/threatened, not yet materialized). Anyone who wants to claim that violence is bad has to oppose the existence of states.
Violence is risky, dangerous, unpredictable, costly, etc. But those are practical and legal, not moral, concerns.
Violence is also necessary, as you say, against governments or other actors which cannot be deterred, stopped or punished using other means.
Violence is also most effective when it's certain and overwhelming/indefensible. If we lived in a world where dictators and their flying monkeys get regularly shot or droned to death, we wouldn't have dictators. Not because they'd all end up dead but because nobody would dare try becoming or supporting one.
This is why we have to publicly support _proportional_ punishment of dictators and their supporters, both now and after they've been removed from power. Good people have to use the same tools as bad ones (after all, they are just tools, not good or bad).
> Don't say sadly...It is a tool, it cannot be good or bad
It's not just a tool, it's also a human action. An action that exacts consequences on its victim and its wielder. Necessary and regrettable aren't exclusive.
On the contrary, target is a neutral word, justifying the violence against a target is exactly as difficult as it should be - based on the circumstances instead of emotions.
Victim already implies wrongdoing so it makes justifying just violence harder than it should be.
For the record, i often use target instead of victim when talking about harassment, bullying, rape, etc. because it also doesn't imply surrender to the aggressor or lack of agency.
The problem is that it is routinely misused (especially by those who have overwhelming power), and the cases where it is really needed are really, really, really rare.
Even in cases when it appears that the use of violence is justified, the long term consequences (e.g. on culture and mentality, and hence ultimately on normal daily life) are usually such that it would have been better to avoid it in the first place.
At the moment you regularly shoot/drone the dictators, the one deciding who is dictator warranting such violence is the most scary dictator of all.
This talk about good/bad people is such naive childish ploy, are we adults here or what?
And that's why it's important to establish publicly known and accepted rules about this. Nobody suggests one person deciding this, usually people who imagine this situation have some issues of their own.
But the threat of absolutely any citizen having a decent chance of successfully killing a dictator would probably lead to democratization of power - individuals would not be attracted to having so much power they would likely become targets and we'd hopefully see more effort towards establishing more direct means of decision making.
> This talk about good/bad people is such naive childish ploy, are we adults here or what?
No need for insults, it's a simplification. It's obviously a spectrum. But broadly speaking, people who regularly intentionally harm others for their gain or pleasure (or see nothing wrong with doing it or support those who do) are considered bad. People who go out of their way to help others are good. The rest is neutral. Most people are neutral - don't see injustice or wrongdoing as their problem until it directly negatively affects them.
And obviously, there are people who do both a lot of good and a lot of bad. I consider those bad because more often than not, they only do the good things to gain support or compensate so they can get away with the bad things.
That's my personal opinion and experience. Other people could for example argue for simply summing up the good and the bad and the total would neatly categorize them. Intent also matters and that's even more complex but usually unprovable from the outside.
Some tools are definitely better than others. Also some tools are not "the right tool" for the job.
Fundamentally though I'm not sure I agree with you. Violence is often an emotional reaction. When violence is used as a tool it is usually (always?) used by bad people.
If it helps you reconcile my worldview, I absolutely oppose the existence of states.
Keep in mind this needs to be judged separately in the legal, practical and moral dimension. For example a state might determine that a person _legally_ deserves to spend 10 years in prison. But the same state will attack you in turn if you abduct that person and hold them for 10 years in similar conditions to prison because _practically_, it weakens the state's monopoly on violence, even if _morally_ that action can be justified (i.e. because if a punishment is just there is no moral reason why it should matter who carries it out).
> often ... usually (always?)
I think the crux lies in how we quantify this. If you live in a western democracy, almost all of the violence you come into contact with or hear about is in fact used by bad individuals (thiefs, gang members, drunks, etc.) or the mentally ill. But even then you have the right (moral and usually legal) to defend yourself.
If you live in other places, that violence might more often be used be institutions (such as states or religions) and might not be materialized (it is potential/threatened/implied). E.g. what happens to a muslim woman who refuses to cover her face - the violence usually never happens because she knows it would and therefore doesn't break the rule. It is still violence used to achieve a goal though and she has the same (moral but usually not legal) right to defend herself - even if any practical defense is beyond her ability to do so because the aggressors are too numerous and dispersed.
I would argue that billions of people live in countries where violence is used against them every day because it is a threat which for example stops them from freely accessing information.
In that regard you're right that it is usually used by bad people. But it says nothing about its morality. The way I see it, violence being used by bad people is a stable equilibrium but it can be used by good people during a transition to a different stable state. It is usually not used by good people in a prolonged because materialized violence tends to reduce the number of people on both sides and cannot be sustained forever.
I don't believe that punishment can be just, and ergo I don't see a moral axis on which to judge violent actions for this purpose. I might concede the use of threatened violence as a means of control, but I don't see any pragmatic way to accomplish this without at least occasional actual punishment so it's a bit non sequiter.
It's likely our views are divergent enough that we wont come to a consensus on this, but I appreciate the nuanced discourse!
I've encountered such opinions before but never cared to engage with them since they seem utterly alien to me. Can you give me a summary of your opinions or links to some materials?
There are multiple goals to a punishment - e.g.:
- Deterrence
- Protection of others / prevention of re-offense
- Removal of aggressor from community to minimize further trauma for the victim by having to interact with him
- Restitution
- Retribution
- Vindication
- Removing any gains from the offense from the aggressor
- Further disadvantaging of the aggressor to make up for `expected_gains * probability_of_getting_caught`
- Further disadvantaging of the aggressor to put negative evolutionary pressure on such behavior
- Separation of the aggressor from others to prevent him from normalizing / spreading his behavior
These few are just off the top of my head, not all apply to all offenses, and not everyone will agree all of these are desired by their favorite society. But how do you achieve any, let along most, of them without punishment?
Sure. It's not something I was formally taught, but rather a viewpoint I came to over time, informed by my own anecdotal experiences and observations. So I don't have any links to share with you, but I'm happy to give a go at explaining my logic.
The naive summary would be that if an individual harms another this is "bad" (a harm to society). To then enact a punishment would be to apply another harm, which would be doubly bad. In my mind, justice is the restoration of balance to the whole, so double harm can never be considered just.
I understand the "goals of punishment" you've outlined and don't even disagree that they can be met (to some extent) via this mechanism. Though I wouldn't classify restitution as a "punishment"[0]. I can also see how the harms from punishment can be moderated when enacted by a state (a group of individuals holding some claim to legitimate power). The question is not whether a system of punishment is sufficient, but whether it is necessary.
If the goal is to reduce all harm to the minimum practical amount we must be pragmatic in our analysis and execution.
I have also seen states grossly amplify the harms involved well past what any individual could be capable of and I have seen individuals operating on behalf of the state launder harm through their office to absolve themselves of any personal accountability for what are ultimately personal actions.
From my observations, those who receive punishment after causing harm rarely become reformed and productive members of society. If anything the punishment[1] leaves them bitter and often actively disadvantages their opportunities for success, leading them to commit more harms and drawing us down a dangerous spiral. Even in confinement these individuals often commit harm to their peers or to those that guard them—and, more abstractly, to the tax payers that are compelled to back this system.
It's not that I doubt these methods can "work" to some extent, but that I see us stuck in a local maxima. I believe we can do much better.
I believe the vast majority of crime that is commited in modern societies could be eliminated through better education, cooperation, and social equality. Why commit crimes, especially violent ones, when the opportunity cost of not pursuing more productive activities is large?
For what remains, I would propose applying insurance. When the volcano erupts and the magma damages my house I do not feel the need to "punish" the volcano.
I do believe there is a floor to violence (and other criminal acts) that will exist no matter what we do. (One-off crimes of passion or acts commited by the criminally insane, for instance). For some of these there may not be any better answer than imprisonment or exile in order to protect the community from future acts. However, I'd love to explore the proactive solutions I recommended above (education, equality, etc) and leave the shrinking vestiges of the current system to handle these edge cases if necessary.
[0]: Likewise, I can see the utility of—e.g—restraining orders and would not call those a "punishment" either, though in this we start to get into the grey area of restraint (limiting an individual's freedom of movement could be classified as a harm).
That's a good way to think about it but unfortunately, human language is so imprecise that IMO many people will leave with the conclusion that "sadly" means "using violence makes me sad and implicitly is therefore bad".
Ideally we'd live in a society where laws are a complete and consistent description of a valid (also complete and consistent) moral system. That's not the reality.
(If it's possible at all because morality operates on reality while legality operated on provability - a subset of reality which can be proven to a neutral third party.)
I suspect this kind of nuance is lost on the sort of people who think having qualms about the use of violence is the same thing as pretending to be saintly pacifists.
Why would you want to, unless you live in a domain of indoctrination ("echo chamber") that pacifism is good and anything else is bad?
I always find it useful to ask "why", whenever someone tells me their beliefs. Children do it and adults sometimes tend to find it annoying because they realize they cannot justify their beliefs but being children, they are easy to dismiss. Harder to dismiss an honest question from an adult.
“If we lived in a world where dictators and their flying monkeys get regularly shot or droned to death, we wouldn't have dictators”
While I agree with the sentiment, the groups who support dictators (oligarchs, religious extremists) would decide to also use violence. So both dictators and the leaders on the side of the people would be murdered and society would be destabilized.
We need reliable anonymous communication as yet another source of friction (drink!) which the state needs to overcome to subjugate the people. And that's why so many states, even western democracies, are trying to oppose it now using children or terrorists as an excuse. The authoritarians and wannabe-dictators (most of whom will never achieve their goal or even publicly state it) are already in government positions, always have been.
There are two upsides:
- There are more normal (good or neutral) people than there are authoritarians (bad people - who want to exercise unjustified control over other people's lives). If the leadership attributes are evenly distributed, then they need to kill more of us than we need to kill of them.
- I don't think people should need to be led. It's a symptom of submissivity many have been taught since childhood ("do what I say and don't talk back") and to some extent is it probably natural but hopefully it can be reduced through better upbringing. Teach your children to question everything and to guess people's incentives and motives. What we need need is enough independent thinkers who are able to communicate and self-organize.
Using violence against someone is the ultimate authoritarian act, so for one side this is business as usual while for the other this is the epitome of hypocrisy.
Your mention of anonymity reminds me of assasination politics [0], which is an idea I found enticing in the past. However I've since come to the opinion that such a system is neither optimal nor necessary, though I believe a similar outcome may be inevitable as we continue along the arc of the democratization of power through technical proliferation.
Only in single-step moral systems (one which judges actions as moral or immoral solely based on those actions in the moment and not what preceded them).
I have a multi-step moral system. Basically any unjustified intentional harm to a person justified proportional retaliation. Unjustified means it is not harm which is being caused as punishment to a previous offense. And proportional means that it shouldn't be too weak, neither too strong. IMO the optimum is causing something like 1.5-3x more suffering/"disadvantagement". However, it is important to signal to both the original aggressor and any potential witnesses why this is being done so that one is not mistaken for an original aggressor himself.
I am also a fan of judging others by their own moral principles. Basically, if someone thinks it is OK to, for example, limit my freedom or harm me (for various reasons or in various circumstances), I apply the same rules to him and it is therefore OK for me to limit his freedom or harm him (for similar reasons or in similar circumstances).
Either system leads to similar outcomes. (The first allows stronger response to offense, the second allows only mirroring).
Thanks for the link, it looks very interesting but it goes into my to-read list for today.
> seems to only have a good track record in places with a democratic tradition
"All that said, there are very obviously regimes in the world that have rendered themselves more-or-less immune to non-violent protest. This isn’t really the place to talk about the broader concept of ‘coup proofing’ and how authoritarian regimes secure internal security, repression and legitimacy in detail. But a certain kind of regime operates effectively as a society-within-a-society, with an armed subset of the population as insiders who receive benefits in status and wealth from the regime in return for their willingness to do violence to maintain it. Such regimes are generally all too willing to gun down thousands or tens of thousands of protestors to maintain power.
The late Assad regime in Syria stands as a clear example of this, as evidently does the current regime in Iran. Such regimes are not immune to an ‘attack on will,’ but they have substantially insulated themselves from it and resistance to these regimes, if it continues, often metastasizes into insurgency or protracted war (as with the above example of Syria) because the pressure has nowhere else to go" (Id.).
> normal Iranian people who just want to leave their life?
Like the ones who are protesting? Idk, when people put themselves in front of a gun I'm inclined to listen to what they're demanding, not folks in their armchairs a world away.
Hitler was so bad that anybody is willing to publicly talk about killing him, there are movies glorifying it, people talk about going back in time and killing baby Hitler. He was so bad that the very strong taboo against killing does not work on him.
So, when _exactly_ did it become OK to kill him? Think about it.
What cumulative sum of his actions between 1889 and 1945 tipped the balance?
Now, do those same rules apply to current dictators or people in the process of becoming dictators even if the taboo is still strong there?
Are you comparing Iran to Hitler?! That does not make sense whatsoever.
If you mean 'At some point, you have to step in and make the change by force. Like we did with Hitler'.
I will say: Yes, at some point it is justified to step in. But, there must be a realistic chance that you will make things better, and low chance that you will make things much, much worse. International consensus would be highly desirable, as well.
In case of Iran: How sure are you that you can make a positive change in Iran by bombing only? If you kill (directly or indirectly, e.g. starvation/ruined water supplies) much more Iranians than Iranians killed themselves (like we did with Saddam), are you really helping?
I don't think there is a will (and maybe not even a capability) for boots on the ground. So, you are just hoping that the new regime would be a better one. Not many positive historical examples there.
Last, but not least: There are serious escalation dangers. What if China/Russia provided Iran with targeting data and/or missiles (not that Iran does not have their own) and Iran hit/sunk a carrier and some destroyers? Are you now in war with China/Russia? At what moment do you cross that line? Will you retreat with the tail between you legs, like from Afghanistan? Or will Israel decide to toss a nuke or two?
The idealism of helping the poor protesters is a noble one, but the road to hell is paved by good intentions.
IM messages aren’t really documents. They are text with some very minimal formatting that could be expressed with markdown. Any media attached isn’t embedded in the document, it’s attached externally / rendered at the bottom.
The only example I can think that messages are expressed as documents is Microsoft Teams. And it’s as much an example of what not to do as anything.
I'd disagree with that for most messaging apps. If you think about Discord or Slack for example. You have a plain text message and then media attachments externally. This could be very well expressed with JSON.
Very few messaging apps let you go beyond plain text and let you start embedding media or complex layouts inside a message.
Slack messages have a ton of formatting. You could implement it with some sort of extension on markdown but you'd have to write a custom parser. XML gives you a markup structure for free.
Slack canvases have full layouts including images.
Eh, XML is a machine-readable generic markup language. Why would you prefer using a less powerful format like markdown in a context like message representation? XML with inline tags seems the perfect fit.
Less powerful also means less complex and less exploitable. You can very easily grab a markdown renderer rather than trying to decode a .docx for messages.
Pretty much no messaging apps let you create messages more complex than markdown anyway.
I think this wave of guilt-by-association is starting to go a bit too far. There were some douchebags around Epstein, for sure. But knowing a douchebag, taking donations from a douchebag, or even going to parties hosted by a douchebag does not automatically make you a douchebag. Social people go to a lot of parties, and money is just money.
Call out the miscreants and bad behavior in the Epstein files, sure. But be specific. I don't like these articles insinuating "the world is full of douchebags". It isn't.
Oh please, the actual files have these people not just vaguely knowing Epstein, but asking him for help with dating, using him as advisor in dealing with sexual harassment issues in their schools (!), actually helping him to get further connections, connecting him to girls, praising him for intelligence in public and so on and so forth.
Epstein credit with them went up after he was convicted, not down. They were perfectly fine with who Epstein was and what he was doing.
Epstein thing did not went far enough at all. Instead, the actors are being protected, circling the wagons against each other.
I feel like I'm taking crazy pills when reading all of these "what is the big deal about being in the Epstein files" takes.
Like, there are people asking for Epstein's advice on how to fuck women that they have huge positions of authority and saying that she is "doomed" sleep with them. And these people haven't lost any material status yet! There are people asking Epstein for help suing feminists who sought to share stories of workplace sexual harassment. And there are tons of people who have publicly said "I was never friends with Epstein" all but making doe-eyes at him in their emails.
And suddenly we have a raft of "oh is it bad to have friends" and "don't all men want to fuck 16 year olds" articles coming from all corners. Insane.
Replace "douchebag" with what you should have written, namely "convicted child sex trafficker, rapist, drug and weapons dealer, torturer and sexual predator", and nothing what you said is anywhere near acceptable.
Besides, the article does not insinuate that "the world is full of douchebags"; it claims that a surprising proportion of people in a very specific, small subset of all people—male authors of a certain age connected to the Edge Institute—apparently had no qualms associating themselves with a person of the aforementioned description. You make it sound trifling when you call it "bad behavior".
I think the fallacy at hand is more along the lines of "no true scotsman".
You can define understanding to require such detail that nobody can claim it; you can define understanding to be so trivial that everyone can claim it.
"Why does the sun rise?" Is it enough to understand that the Earth revolves around the sun, or do you need to understand quantum gravity?
Good point. OP was saying "no one knows" when in fact plenty of people do know but people also often conflate knowing & understanding w/o realizing that's what they're doing. People who have studied programming, electrical engineering, ultraviolet lithography, quantum mechanics, & so on know what is going on inside the computer but that's different from saying they understand billions of transistors b/c no one really understands billions of transistors even though a single transistor is understood well enough to be manufactured in large enough quantities that almost anyone who wants to can have the equivalent of a supercomputer in their pocket for less than $1k: https://www.youtube.com/watch?v=MiUHjLxm3V0.
Somewhere along the way from one transistor to a few billion human understanding stops but we still know how it was all assembled together to perform boolean arithmetic operations.
With LLMs, The "knowing" you're describing is trivial and doesn't really constitute knowing at all. It's just the physics of the substrate. When people say LLMs are a black box, they aren't talking about the hardware or the fact that it's "math all the way down." They are talking about interpretability.
If I hand you a 175-billion parameter tensor, your 'knowledge' of logic gates doesn't help you explain why a specific circuit within that model represents "the concept of justice" or how it decided to pivot a sentence in a specific direction.
On the other hand, the very professions you cited rely on interpretability. A civil engineer doesn't look at a bridge and dismiss it as "a collection of atoms" unable to go further. They can point to a specific truss and explain exactly how it manages tension and compression, tell you why it could collapse in certain conditions. A software engineer can step through a debugger and tell you why a specific if statement triggered.
We don't even have that much for LLMs so why would you say we have an idea of what's going on ?
It sounds like you're looking for something more than the simple reality that the math is what's going on. It's a complex system that can't simply be debugged through[1], but that doesn't mean it isn't "understood".
This reminds me of Searle's insipid Chinese Room; the rebuttal (which he never had an answer for) is that "the room understands Chinese". It's just not satisfying to someone steeped in cultural traditions that see people as "souls". But the room understands Chinese; the LLM understands language. It is what it is.
[1] Since it's deterministic, it certainly can be debugged through, but you probably don't have the patience to step through trillions of operations. That's not the technology's fault.
>It sounds like you're looking for something more than the simple reality that the math is what's going on.
Train a tiny transformer on addition pairs (i.e i.e '38393 + 79628 = 118021') and it will learn an algorithm for addition to minimize next token error. This is not immediately obvious. You won't be able to just look at the matrix multiplications and see what addition implementation it subscribes to but we know this from tedious interpretability research on the features of the model. See, this addition transformer is an example of a model we do understand.
So those inscrutable matrix multiplications do have underlying meaning and multiple interpretability papers have alluded as much, even if we don't understand it 99% of the time.
I'm very fine with simply saying 'LLMs understand Language' and calling it a day. I don't care for Searle's Chinese Room either. What I'm not going to tell you is that we understand how LLMs understand language.
Your ultra-reductionism does not not constitute understanding. "Math happens and that somehow leads to a conversational AI" is true, but it is not useful. You cannot use it to answer questions like "how should I prompt the model to achieve <x>". There are many layers of abstraction within the network - important, predictive abstractions - which you have no concept of. It is as useful as asking a particle physicist why your girlfriend left you, because she is made of atoms.
Incidentally, your description of LLMs also describes all software, ever. It's just math, man! That doesn't make you an expert kernel hacker.
It sounds like you're looking for the field of psychology. And like the field of psychology, any predictive abstraction around systems this complicated will be tenuous, statistical, and full of bad science.
You may never get a scientific answer to "how should I prompt the model to achieve <x>", just like you may never get a capital-S scientific answer to "how should I convince people to do X". What would it even mean to "understand people" like this?
No one relies on "interpretability" in quantum mechanics. It is famously uninterpretable. In any case, I don't think any further engagement is going to be productive for anyone here so I'm dropping out of this thread. Good luck.
Quantum mechanics has competing interpretations (Copenhagen, Many-Worlds, etc.) about what the math means philosophically, but we still have precise mathematical models that let us predict outcomes and engineer devices.
Again, we lack even this much with LLMs so why say we know how they work ?
Unless I'm missing what you mean by a mile, this isn't true at all. We have infinitely precise models for the outcomes of LLMs because they're digital. We are also able to engineer them pretty effectively.
The ML Research world (so this isn't simply a matter of being ignorant/uninformed) was surprised by the performance of GPT-2 and utterly shocked by GPT-3. Why ? Isn't that strange ? Did the transformer architecture fundamentally change between these releases ? No, it did not at all.
So why ? Because even in 2026, nevermind 18 and 19, the only way to really know exactly how a neural network will perform trained with x data at y scale is to train it and see. No elaborate "laws", no neat equations. Modern Artificial Intelligence is an extremely empirical, trial and error field, with researchers often giving post-hoc rationalizations for architectural decisions. So no, we do not have any precise models that tell us how a LLM will respond to any query. If we did, we wouldn't need to spend months and millions of dollars training them.
We don't have a model for how an LLM that doesn't exist will respond to a specific query. That's different from lacking insight at all. For an LLM that exists it's still hard to interpret but it's very clear what is actually happening. That's better than you often get with quantum physics when there's a bunch of particles and you can't even get a good answer for the math.
And even for potential LLMs, there are some pretty good extrapolations for overall answer quality based on the amount of data and the amount of training.
>We don't have a model for how an LLM that doesn't exist will respond to a specific query.
We don't have a model for a LLM that does exist will respond to a specific query either.
>For an LLM that exists it's still hard to interpret but it's very clear what is actually happening.
No, it's not and I'm getting tired of explaining this. If you think it is, write your paper and get very rich.
>That's better than you often get with quantum physics when there's a bunch of particles and you can't even get a good answer for the math.
You clearly don't understand any of this.
>And even for potential LLMs, there are some pretty good extrapolations for overall answer quality based on the amount of data and the amount of training.
> We don't have a model for a LLM that does exist will respond to a specific query either.
Yes we do... It's math, you can calculate it.
> No, it's not and I'm getting tired of explaining this. If you think it is, write your paper and get very rich.
Why would I get rich for explaining how to do math?
> You clearly don't understand any of this.
Could you be more specific?
Quantum physics is stupidly hard to calculate when you approach realistic situations.
A real LLM takes a GPU a fraction of a second.
They're both hard to interpret, please realize I'm agreeing that LLMs are hard to interpret. But they're easier than QM on some other fronts.
And mentioning copenhagen or many-worlds doesn't show that quantum mechanics are easy to interpret, that's about as useful as saying an LLM works like neuron activation.
It would explain why the information density is so low. I got about halfway through hoping for something clever, then started scanning, then just gave up.
10k words for "migrating code and data is a headache". Yeah. Next.
reply