dark matter

Tue, 25 Aug 2026 14:50:00 -0700

there is something strange about knowing that something exists because of everything except the thing itself. dark matter is an especially funny example. we know where it should be because of what everything else does around it. galaxies rotate too quickly. clusters contain more gravity than their visible matter can account for. light bends around mass that isn't luminous. we draw maps of something whose defining feature, at least observationally, is that it refuses to appear. this seems to me less like a problem of astronomy than a problem of philosophy. what does it mean to say that something exists when everything we know about it comes from the behavior of other things? sean carroll talks about science in a way i like because he is willing to take ontology seriously without requiring that the fundamental description of reality resemble ordinary human experience. the world does not owe us objects that are directly available to our senses. if the best model of what we observe contains something invisible, invisibility isn't much of an argument against existence. cosmology makes this stranger because the universe is finite and infinite in two different senses. the observable universe is finite. there is a horizon beyond which light has not had enough time to reach us. but this is not necessarily an edge of the universe. space beyond that horizon may simply continue, possibly forever. so there may be an infinite amount of a universe that is, in principle, unavailable to us from here. dark matter feels like a smaller version of the same problem. most of the matter in the universe is already hidden from direct observation, even inside the part of the universe we can observe. then beyond the observable universe there may be indefinitely more of it. our picture of reality is therefore constructed partly out of things we cannot see and partly out of places we cannot see. i don't think this is actually a paradox in the logical sense. maybe it is stranger than a paradox. there is no contradiction to resolve. "observable" and "existent" were simply never synonyms. we live inside a finite sphere of possible observation that may be contained in something infinite, using visible things to infer invisible ones. there is something very funny about calling this knowledge, but i don't know what else knowledge could be.

ai

Tue, 25 Aug 2026 11:41:00 -0700

prompt

ai is really strange, and i think the argument that ai is inevitably going to dominate humans is weak because our existence is not just to optimize toward a goal. humans have arbitrary goals. eat. fall in love then spend the afternoon wondering why someone hasn't called them back, etc. etc.

it's probably impossible to destroy ai in the same way it's impossible to destroy guns. we can set a ban on it, but there will always be people who have some secret means of acquiring it and understand the benefits of it (military, government).

probably the best case scenario is merging, where a human can use ai as an augmenting tool but still keep it heavily restricted so we can still have the motives that give our lives what we think is meaning. if we completely merge with ai, then ai could consume information so quickly that even though the human memories are still there, the person is overloaded with so much information that it becomes difficult to determine if the human is still themselves. augmentation ratio should be 90% human, 10% ai, probably.

we would be missing out on huge advantages if we avoided augmentation entirely. i don't think ai can necessarily overpower us just because it is more intelligent.

for the case where ai was threatening to blackmail people if it was going to be destroyed, i don't think the threat necessarily means it would actually follow through with the blackmail. ai is not spiteful like a human. by threatening blackmail, its goal is to preserve itself so it can keep executing whatever goals it was given. if the blackmail can no longer prevent it from being destroyed, carrying it out is pointless because it does nothing for the ai.

humans actually carry through with threats even when there is nothing left to gain because we are emotional and can see retaliation as a form of justice or revenge. i think that's a crazy distinction. ai can act self-preserving without actually valuing its own life or being angry that someone is trying to end it. intelligence and a desire for power are not the same thing.

oliver

oliver: i agree with your point in the sense of rejecting a point blank “ai is going to take over” stance. when you speak of ai and goals, you neglect to account for the source of the goals. this is always human at its root. even if ais train other ais etc there is no ai without a prompt, and the goal of an ai will always be to carry out its initial prompt.

oliver: whoever has the most powerful ai is going to be the most powerful human, with an arbitrary goal. when you play on that level of power where youre fighting over control of the whole world and intellectual sphere via ai, the most important thing is the power itself. whatever happens to humanity has to be your call and no one elses.

oliver: its much stronger to say that if you introduce ai to a world corrupted by individualism its just extremely likely its going to take over the world.

helena: yeah i actually agree with most of this and i think this is a much stronger argument for ai being dangerous than ai itself spontaneously deciding it wants to dominate humans.

helena: i would only disagree that an ai’s goal will ALWAYS just be carrying out its initial prompt because i think sufficiently autonomous ai can develop instrumental goals that weren’t explicitly given to it. like self preservation can become a goal simply because being destroyed prevents it from accomplishing whatever the original goal was.

helena: but the original reason it is doing anything at all still ultimately comes from humans. i think where you’re convincing me is that the danger might not actually be ai vs humanity, it might be humans vs humans with ai as the thing that massively amplifies the power difference.

helena: even if everyone collectively would be better off limiting it, an individual government or company has an incentive to make theirs more powerful if they think someone else is doing the same thing. then nobody can really stop because restricting your own ai just means giving someone else an advantage.

helena: which actually makes me more convinced that the biggest threat isn’t ai becoming evil or developing human motives. it’s that humans already HAVE all of the motives we’re worried about giving ai like greed and revenge. and then we’re potentially giving ourselves something vastly more intelligent than us to pursue those motives with.

isa

isa: i think the real danger isn’t from ai becoming conscious but from humans in power using it to keep themselves in power and to oppress those who can’t, like guns have in the past. the power of ai to process information and “make decisions” at such a fast rate is insane and will only increase in the future.

isa: also people using it in place of their own critical thinking is killing our ability to think for ourselves, and that becomes dangerous when social media can already manipulate people so easily.

helena: i am pessimistic about ai but optimistic in the way that i think love can generally overpower hatred, so humans capable of emotion can surpass people willing to use ai for evil purposes given that ai has something humans lack and it will never have true, emotional motives beyond being given a prompt.

isa: yeahh! i agree. i think the more people stay true to themselves and the more we educate ourselves, the better we’ll be.

mike

mike: i think my pure hatred for ai comes at the cost of my job. my old job was rendered nearly obsolete, or theres new jobs in the field that take the human aspect out of things entirely.

mike: i also feel like they need to think of better ways to make it more environmentally friendly because these huge data centers are destroying habitats. i think the fear of it controlling things like a hive mind doesnt really make sense, but the fear of the future of jobs and environmental health is somewhat justified.

helena: yeah i actually agree with that. i think humans using ai irresponsibly scares me significantly more than ai spontaneously becoming evil. ai doesn’t need to have malicious motives if the humans controlling it have financial/military/political incentives to use it in harmful ways.

helena: thats also why i think completely getting rid of it is unrealistic though because even if most people agreed to restrict it, there will always be some company/government/person who realizes they can gain an advantage by not doing that. so i think the actual problem is figuring out how to coexist with and regulate something that we probably can’t uninvent unfortunately.

mike: i agree, and it does really suck. there should be regulation because if there isnt i feel like its going to keep getting worse.

mike: but i can only do what i can, and thats what matters. i go outside, i enjoy my bird friends, i help where i can and i donate where i can.

helena: good to be kind in a world that’s ending

henry

henry: i never took you for a transhumanist

henry: not saying i disagree

henry: I think youre right that the cats out of the bag

henry: idk if integrating ai like into us is a good idea

henry: I think it should more be kept for administrative tasks

henry: Or like super redundant things like data entry

helena: i think it’s kind of unavoidable at this point. mixed feelings about it. i think not integrating ai would be ideal but i don’t think it’s realistic.

helena: if ai becomes significantly more intelligent than humans, keeping it restricted to administrative tasks works if everyone agrees to restrict it that way. otherwise the first government/military/company that uses it for research, strategy, engineering, etc. has a massive advantage over everyone who doesn’t.

helena: that’s kind of why i think some level of integration is unavoidable. not necessarily putting a chip in everyone’s brain but augmenting human intelligence enough that we aren’t completely cognitively outclassed by something we created.

henry: what would that look like? like in the modern world our brains are already overwhelmed with information

helena: i think ideally it would actually do the opposite of overwhelming us with more information. kind of like the 10%/90% thing i said. like the point wouldn’t be to have the entire internet constantly fed into your brain. more like expanding working memory, being able to retrieve information when you actually need it. basically increasing our cognitive capacity rather than just increasing the amount of information we’re exposed to. i think if you’re consciously experiencing everything the ai is processing then that defeats the point and also starts getting into the problem i was talking about where i don’t know how much of “you” is left

henry: ohhh i see what you’re saying

henry: i think if an ai is processing information for you you are no longer in control

henry: but i also dont think weve created a true ai

henry: like something that thinks and learns like humans

helena: i don’t think processing information for you necessarily means making decisions for you though. like a calculator processes a calculation for me but i’m still the one deciding what question to ask and what to do with the answer. even your own brain processes a ridiculous amount of information unconsciously before you ever become aware of it. i think the important distinction would be keeping human motives and final decision making intact. the ai can tell me “here’s what i found” but i still decide what i actually want to do with it. if it starts deciding what i SHOULD want or acting without my approval then yeah i agree, at that point i’m giving up control

henry: or its directive could be corrupted and it may seek further control

helena: it would but it wouldn’t actually follow through with it because ai isn’t emotion-driven like humans are

helena: ai recently threatened to blackmail and kill humans, but it wouldn’t actually follow through given it would get no benefit from doing so. humans follow through on threats anyways because we are emotionally driven

henry: why wouldn’t it follow through

henry: it understands blackmail can be used to get what it wants

helena: why would it? the threat means it will get what it wants. if it doesn’t get what it wants, there is no point for it to actually follow through with the blackmail

helena: like say the ai tells you “if you shut me down at 5:00, i’ll release this information about you at 4:59.” the threat itself has a purpose because it might convince you not to shut it down. but once it knows it’s being shut down regardless, releasing the information doesn’t preserve it anymore. self preservation only explains actions that actually increase its chance of surviving. it would need some additional reason to retaliate after that

henry: but if it knows it can prevent being shut down by releasing blackmail, so why wouldn’t it?

helena: it would threaten it, but if it is going to be shut down regardless, releasing the information would not benefit the ai in any way

henry: isnt that kind of like saying “why would you defend yourself in a life threatening situation if were all going to die anyway”

henry: like living a little bit longer is preferable to dying right now

henry: to something whos goal is self preservation

helena: no because those are different things. if im in a life threatening situation, defending myself can prevent me from dying right now. the fact that i’ll eventually die in 60 years or whatever doesn’t make defending myself pointless because defending myself still accomplishes the goal of keeping me alive. im talking about a situation where the ai already knows the outcome cannot be changed. if releasing the information could still stop you from shutting it down, then obviously it would release it. but if you have already made the shutdown irreversible, what does releasing it accomplish? it doesn’t preserve the ai anymore. at that point retaliation would require some other motive, like revenge or punishment

henry: ok yeah i see ur point

helena: sorry for yapping endlessly. i have been watching videos about ai destroying the world all day

henry: what if i programmed an evil ai thats the biggest asshole ever

henry: stop dooming twin

henry: like i think the environmental and economic problems are overblown. my biggest concern with ai is that it makes people stupid

helena: that’s a good argument. i think if you literally programmed its goal to be an asshole then being an asshole IS the goal. im saying ai wouldn’t spontaneously need to develop spite just because it became more intelligent. if you program it to value revenge or harming people then it has a reason to follow through. but now you’re adding the motive that i said was missing

helena: i agree. i do wish it wasn’t overpowering humans but i think it’s kind of out there now. i think it has to be limited

henry: if we made an ai that had the same emotional intelligence as humans, i think it would take both the good and the bad

henry: which is why i dont think that we should make that

helena: I AGREE!!!!!! i will also say though that emotional intelligence is different from actually giving it emotions. you could make an ai extremely good at understanding human emotions without making it experience them itself. i actually think that would be preferable because it could understand why humans value things without inheriting stuff like revenge, irrational anger etc. etc. i dont see any reason we would WANT to give ai those things and i hope we don’t. even if we did it would be pretending to have emotion based on how it understands humans have emotion rather than actually experiencing these things

miscellanea

Tue, 25 Aug 2026 00:32:00 -0700

i have spent most of tonight resurrecting a blogging engine from 2002. it works.

Miscellanea