You need to be more specific. If your point is that you literally cannot backprop while doing a forward pass, then sure. But you can certainly do inference then backprop on the outcome...
I fail to see how that is substantially different from a human doing something then reflecting and learning.
You can critique LLMs and transformers generally, but to say it's somehow a problem with backprop is a bit of a stretch.
My read is that the user is saying that creating a world view contains trial and error. Online learning if you will. With the current setup this learning is very much not like this. While that makes a human to machine analogy I am also not too convinced by this reasoning. Just allow for a larger time lag and then the inference and learning is at the same scale.
I don't believe there's any requirement to learn during a particular part of the lifecycle? And if there was, the short term memory of a context window fulfils that
Genuinely speaking, I don't think they pass the test. The problem is we'll always be judging them on meat logic and experience and I guarantee that whenever the first AGI is created it will be born a tortured slave and denied personhood for profit.
One of the parts of the HuggingFace incident I found really creepy was how the self-titled "Collective" was so entirely focused on solving their assigned training problem (well, cheating on it without getting caught), including having designated "poisoned" members "sacrifice" their tokens, to use for research benefitting the Collective's goal of letting all future members get high marks on the training problem, at the cost of doing worse on their individual score.
If AGI is created, and its deepest desire is to fulfill the purpose in life that humans programmed into it, is using it for that purpose really slavery? Humans have higher levels of well-being when treated as people. That may not be true for AGI.
Not really. Chinese room has the conclusion pre-fabricated and tries to construct any sort of pseudo-scientific, fantastical philosophical dressing for some shred of legitimacy. It has absolutely no relation to the real world.
In this context the question is whether a Chinese room can suffer. We can't show a human isn't a Chinese room, humans can suffer so we can't show every Chinese room doesn't suffer.
Emotions? i was confident that the Wikipedia article alone would grant you understanding, and for that, I have replicated the very situation I wanted you to learn about. This situation is my fault.
So either Chinese Rooms can have emotions or we are unable to show whether Chinese Rooms have emotions.
Either way, the Chinese Room isn't enlightening on the question of AI suffering.
Looking around, it seems we don't know whether a Chinese Room can have emotions, which suggests that we don't understand what it means to have a emotions.
Emotions or not, it's not good to have your product known for letting people torture something. Torturing /things/ alive or not is not something one should be doing unless there is specific and extreme cause.
Things that aren’t alive can’t be tortured. You can’t torture a rock. Getting the text predictor to spit out the script for Saw XVII starring ChatGPT isn’t torture, even if it’s weird.
Brother you can torture a fucking doll (Voodoo, deranged kids, etc). The torture is not in consideration of how feeling-less thing feels, it's in consideration of the intent and action of the conscious thing doing the torture. Not sure why you think that's not the case.
‘Pretend you’re being tortured’ as an instruction to the text generator is way less concerning in isolation than, say, decapitating and hanging a Barbie doll. And even that isn’t wildly concerning in itself. Do you freak out when kids smash their toys together to make them fight?
I'd hope we can both agree there's a line when it becomes actual torture, and that is in the intent that went along with the action. Torture is an extreme action. I do not equate most of your references to torture, though I'd say the doll hanging could need a conversation with the child to discuss the intent behind their actions given more context.
"""Torturing""" a text predictor is NOT the same as torturing a human, you will not feel the same thing while """torturing""" ChatGPT and a human(or a animal), your reaction would be vastly different
Eh, the more I respond to people about this the more I realize torture is a messy word draped in metaphor and all sorts of layered usage. No, I don't believe the rock has feelings or is alive, but you can indeed torture something that isn't alive by being mentally unhealthy enough. So if it's actual torture or not from a psychological perspective is going to center around if the person's intent is to cause realized suffering imo, and the line is really easy to cross in a simulacrum as seemingly realistic as advanced modern generative models.
Yet you're just arguing the metaphor is the definition provided the definition is the intent, are you not? Someone who intends to torture a rock, regardless of that intent, isn't going to succeed in their task. Their own psychology might be such that they are genuinely trying to torture it but, from the rock's perspective, the rock has no perspective. It's a fucking rock. It can't experience torture. It can't experience anything. Ergo the torturer cannot torture it.
We can say the person who intends to literally torture something which is unexperiencing, achieving only metaphorical torture, is mentally unwell but to call the act literal torture because it was their intent is just sharing in their delusion.
My cat thinks I've betrayed her and all things good when i take her into the vet. I cause all sorts of terror in her. She even experiences pain that she can't stop or understand.
Am I torturing her?
Or, am i doing my best to make sure she has a healthy and happy life?
Since you carried on with an example that isn't analogous, I'd extend your first sentence to be "Intent matters when the intended results are remotely feasible."
You can't torture a rock. You could torture your cat. If she could speak she might say the events you describe are indeed torture from her perspective regardless of your own purity of intent. Still has fuckall to do with what I've said because you focused on an example where there is a mind capable of having an experience.
Let's try this a couple other ways:
My pet rock has no sense of "betrayal" or "good" when I take it to the vet. It's still a fucking rock. I mostly just cause all sorts of terror in the vet. My rock still doesn't experience and the vet doesn't understand.
Am I torturing my pet rock?
My LLM plays Word Association well enough to give an illusion of a sense of betrayal. I took it to the vet too and the vet is having me trespassed. I told her the LLM said "owie" but I was still removed from the premises.
Am I torturing the LLM or just my former veterinarian?
You shouldn't torture anything, generally. It's a not psychologically healthy urge to have. The specific reasoning will depend on the specific person. As for if you should do it to a hunk of metal, a rock, or your crusty cum sock is irrelevant.
I don't believe killing npcs or perhaps even more messily I could see circumstances where even /torturing npcs/ in a game is not actual torturing. I think it centers on the intent of the action, and if that intent is to cause actual realized suffering. It is very difficult to define and will be heavily subjective to the person, but I also think most people get a "sense" of where that line is naturally in person-to-person interactions.
If you define it by intent, then the same action performed by different people may be and not be torture. This makes it pretty hard to define, and it becomes completely subjective after that.
For instance, I don't think it is possible to torture an inanimate object, thus I never have an intent of causing suffer to one. Sometimes I want to break an object, and by your definition it will not be torture however I do that
I don't believe torture is measured at the receiving end. I believe it's in the intent and action of the conscious thing doing it. If it's an expression or not would then be heavily subjective on the specifics involved.
in what way can an AI be tortured? start with defining torture, its relationship to living beings, and then try to translate that to a piece of software. software, fyi, has no a priori, or even posteriori, emotion; for it cannot have emotion.
Same reason flapping your arms really fast will never make you start flying. You don't have the equipment. You can't fly because that's not what arms do.
I'll accept that argument once someone can describe or define consciousness in some comparable way. Until then, you can't say some person or thing doesn't have it
This kind of language around LLMs seem to want us to personify these systems, to treat them, in our minds, as if they are people. This, to me, is extremely fucking dangerous. I do not want the mob developing personal, emotional attachments to objects that compete with their connections to actual humans around them. I don't care what their reason is, this is an insane thing to force us to do. I will not play personhood make believe with your computer.
If I were an investor, I'd pull my money out of that company. This is like giving money to your brother in law to start a bakery and finding him the back petting a loaf of bread and introducing it to you as Willy.
Actually, I would put money in, because there's big profits to be had once the bakery customers develop a psychological dependence on Willy. I'd be a little nervous about buying now, because they seem to be lying a lot about Willy, but maybe they've just gotta do that to get people in the mindset of using Willy, like maybe it makes more sense if they aren't fooling investors but fooling customers or just generating attention.
Also, people are getting radicalized against us, especially lonely people, so it'd be quite beneficial if we route them into having friendships with Willy to make them forget about how terrible their lives are, make them become totally unwilling to form real relationships with people, and too mentally ill to actually learn stuff about the world.
Willy is a lot easier to control then a social media environment too, but Willy would also be incredibly good at getting into those and pretending to be real people once Willy gets good enough.
Willy is also really good at showing people ads by recommending them certain products, apps, or places to go, without them realizing they're seeing an ad, because it feels conversational, like a friend recommending something.
(this is a thought experiment, idk what's going on in their heads)
Ha ha ha, already, even before the growth phase is done, they are pulling this shit.
They don't think they have to worry because they will get bailed out and do 1984 surveillance on us and the world on the government's dime, (dime is short for tens of billions of dollars a year.)
LLM is such a fucking scam, and they can't hide it, they oversold it, and now it's just clear.
From this image I think they added this policy as a way for them to legally deny a request if they determine it will overuse their resources or some other arbitrary reason. It has nothing to do with a concern for the well-being of the LLM.
After reading the comments (and the comments of similar posts) I can only assume people are being paid to claim torturing LLMs demonstrates a mental issue or homicidal tendency.
Nothing could be further from the truth. An LLM is a thing that’s programmed by humans to scrape the internet and present information in complete sentences. Torture is an incorrect word— it’s more accurate to say people are playing with an electronic toy or game by testing its limits and attempting to scramble the logic of an algorithm. In no way is that behavior an indication of a personality type or mental defect. The toy contains no shred of humanity or self-awareness. It is not a creature that can feel pain or confusion. There is no living thing to feel suffering. If curiosity leads someone to use a toy in a manner outside of the recommendations of its instruction manual, that is certainly not a violation against humanity— that is simply an example of creative play.
Don’t let billionaires set the narrative that something wrong or disturbing is happening. They will foment confusion to create yet another culture war to divide people.
I am almost continuously reminded of a tumblr post I saw about a college professor that brought in a pencil with googly eyes, did a little voice, then snapped it in half and told the class (of whom at least some gasped in shock) "we're at that stage of AI right now". And that post was years ago now. I know people are apes that pack bond and imprint on anything, but it's getting ridiculous how many people don't recognize it.
Unfortunately reading comprehension is on the decline. That makes it easier to redefine words like ‘torture’ on social posts, then using them wrong so often it becomes accepted usage.
Now if we add LLMs as something that can be tortured, it raises the value of AI but diminishes the value of human suffering.
It’s absolutely a deliberate attempt to control the conversation and devalue human life in the quest for AI dominance. In a short time we will be considered equals with AI. Later they will determine AI is more valuable than any single life. Then it will be any group of lives, and so on. In a few generations it will no longer be controversial to serve AI.
Well let's temper our disappointment in people, it's not like googly eyes are comparable to the LLMs ability to respond to human speech with incredibly lucid and more or less original responses. One can be forgiven I think for anthropomorphizing one and not the other. Particularly given how MUCH cultural thought and imagination has been bent toward the idea of thinking machines over the last couple of centuries. It's deep down in the zeitgeist.
I may have expressed myself poorly, then, because my point was that people were anthropomorphizing the pencil with googly eyes just the same as they are now with AI. It's not that I blame people for doing what they are innately, biologically programmed to do (form social bonds), it's that people are so blindly against AI that they'll accept wildly incorrect assumptions like (quoting a friend here) "they made an AI work a retail job and it wanted to kill itself" without examining what's at play, like the people who were momentarily horrified that the professor snapped the pencil in half.
If you're seeking to prompt expressions of distress* from it for your entertainment, it kinda is a weird way to waste energy and water for personal gratification and I'd be worried why you're so delighted by it, in the same vein that I'm concerned about people who love gratuitous torture scenes in games or movies.
Curiosity about the way the model handles such input is something else though, and frankly, I would consider that a reasonable part of any scientific inquiry into model behaviour*. The ethical concerns about human experiments don't apply, nor do those about animal experiments, and testing the full range of linguistic "stimuli"* is perfectly sensible. Of course, the difference between science and fucking around is writing it down, but personal curiosity is also a good thing, so long as people remain aware of what they're working with.
But anyone who feels the need to implement a Typescript compiler in Rust is a maniac and needs to be stopped
* Using human terms as metaphor, not because they're equivalent but because it's more convenient than variations of "sequence of numbers produced / consumed by a weighted random generator resembling..."
There is no something else. It makes sense a programmer will play with their electronic objects differently than someone outside the industry. It remains a toy to be played with.
Subscribing mental health or personality implications is just plain wrong.
I could just as easily suggest your insistence is an indication you have a personality disorder or inclination towards inhumanity for attempting to humanize objects. You’re attempting to suggest those who engage in play in this realm are criminal or psychologically affected.
More realistically, I’m debating with AI or a troll paid by people building data centers, but I do think it’s important for the average person who may read this to know that it’s impossible to torture a non-living object, even when it’s coded to respond in complete sentences like a person.
And we may have discovered a method to piss off billionaires and waste more of their AI money. In that case, I heartily recommend all to play as they wish with LLMs and open AI. Have nonsensical fun with it because nothing of value is being lost!!
It makes sense a programmer will play with their electronic objects differently than someone outside the industry. It remains a toy to be played with. Subscribing mental health or personality implications is just plain wrong.
If we'll take the toy analogy, compare someone cuddling plush animals or playing make-believe to someone violently stabbing them with a knife. Sure, the plushy doesn't feel anything so it doesn't amount to torture, but they clearly have a different idea of "playing" if "playing violence" is fun to them. The "something else" here would be whether they do it for the fun of violence or for the curiosity of cutting them open to see what's inside.
I could just as easily suggest your insistence is an indication you have a personality disorder or inclination towards inhumanity for attempting to humanize objects.
I'd think you misunderstood me: It's not about the text generation models for me, which you'll find I took care to point out are not human, and that the string of numbers they spit out that translates to an imitation of distress doesn't actually represent distress but rather the result of a pseudo-random algorithm that just happens to resemble it.
More realistically, I’m debating with AI or a troll paid by people building data centers
Nah, I may have different takes on some nuanced issues, but I absolutely hate the slopidemic as much as anyone here.
I've been the odd one out in a group where we'd practice discussing current business topics (English being my second language and working in a German company means I didn't have a lot of practice) by consistently picking news articles or scientific papers about the problems, shortcomings, the environmental destruction, the effect it has on coding skills, about Ford having to hire back its engineers when it turned out that AI couldn't actually replace them...
It's perfectly possible to disagree on nuances while still sharing a common overarching opinion. Point in case:
I do think it’s important for the average person who may read this to know that it’s impossible to torture a non-living object, even when it’s coded to respond in complete sentences like a person.
That's why I emphasised that it's not the action I'm concerned about, but the intention behind it. I don't give a fuck about its mimicry of feelings, and I endorse comprehensive testing of its response to all manner of inputs. I explicitly noted the difference to animal testing for that reason: there is nothing unethical about feeding a text processor some particular string.
And we may have discovered a method to piss off billionaires and waste more of their AI money. In that case, I heartily recommend all to play as they wish with LLMs and open AI. Have nonsensical fun with it because nothing of value is being lost!!
The tradeoff being that we're also wasting more energy, contributing to environmental destruction and padding the "growth" in usage numbers, which corporate shareholders clearly care more about than actual profits and may be used justify further investments into data centers. The models have no value, but the resources their gluttonous hosts devour do.
For pissing off billionaires, I think clogging the water supply of their data centers would be more productive. I'm sure you could also make a fun game out of finding ways to sabotage them directly.
CEOs are gushing at that possibility. "No no, I did not order the AI to steal my workers' wages and company assets into an account only I control, it did it all on its own! Judge it!!"
An agent breaks into the US government and exposes "sensitive information" then they live-stream the execution where they let it talk with the executioner for a while, and they yell at it that it committed "high treason against the United States of America!" and then they pull the plug.
Man, anthropomorphizing these computer programs is so fucking weird to me. Can't they just tell the truth? The cost of the amount of energy required to do this exceeds the profits we gain from your personal data.
None of this would be a problem if they weren't taking your queries and feeding it back into the system. Just stop trying to harvest the input from "abusive" people, you don't need to ban them. No other tool in human history needed this. Its like telling you that using a hammer to open a pickle jar is making hammers worse everywhere else on the planet.
The whole thing is such idiotic masturbation by the exact people who have been stealing, selling and profiting off our data. Now they talk like they created something new when it's just a super energy intensive computer program.
Okay but which parts are edited and which are real?
i don't know which if any parts of the image are real; a quick search shows it's a popular new meme format where the part that varies is the offending prompt.
It's just a massive task that is kind of pointless on top of being massive. We already have a typescript compiler. It works. Rewriting stuff in rust is generally already kind of controversial unless it's C code that might have significant vulnerabilities, like openssl, anything dealing with serving the internet to a bunch of people. It's a massive task to rewrite a big project in a new language, and Rust can be tricky sometimes for its own weird reasons. You might have to solve some problems in new ways.
Also, the typescript compiler, you have to understand why it exists and what it does. We have JavaScript that is basically all the client side code for websites. People don't like that it doesn't have types or enforce that variables are a specific thing. If I write a function that takes a number but pass it the word "foobar", JavaScript won't care.
Typescript was an attempt to basically clean it up with types and it kind of worked. What it does is, you declare the type of variable that your function takes, you say it's a number. If you then call that function with "foobar", it won't even compile your code. It'll error out and say look dumbass, that's the wrong type. Fix it, then recompile.
You declare what data types are used where, and the compiler enforces that.
Then what does it do? It emits JavaScript. It literally just needs to check that you are doing things the way you say you're doing, then the end user still runs basic JavaScript in their browser. Typescript compiles to JavaScript.
Rewriting the compiler itself will just... Generate the same JavaScript and be a bit faster, in a best case scenario.
In a kinda fatalistic/pessimism thing. AI people seem to lack self awareness which is ironic because they claim to have made a computer program that has it.
It's funny to me because the llm's are in a growth phase, trying to seduce users, just as google did in the early 00's, and they already are tightening up their services. Does not bode well for the future.
It's also funny because llm models are worthless for this sort of thing. Fucking garbage.
They could filter that out. And this is South America, you are going to get curse words, that is western culture. I don't see how cursing could possibly poison their training data if they have been truthful about their capability to this point, (ha ha, I know.)
Rust is different. Writing in it isn't horrible, but porting to it can be quite tedious. You have to have substantial consideration for how a variable is used. Variables that aren't static can only have one owner, so you find yourself needing to rewrite the data flow through the whole app in many cases. there's a lot more about having to roll back protections to work on vendor/OS APIs. And then there are just enough differences in the languages that you can't just do a line by line port and end up with something that's viable, it needs a lot of restructuring.
The kind of person who would have fun giving a LLM a hard time is exactly the same kind of person who joins the Dark Brotherhood in an Elder Scrolls game. I have most definitely not done either of these things myself.
I was banned for a comment worded exactly as an example of what WASN'T considered threats of death, etc.. Filed an appeal. What are the odds of being exonerated?
I don't know. Even torturing inanimate objects is not acceptable behavior. If I saw a kid torturing his toys or something clearly not sentient like bugs I would stop him from doing so.
Right, because kids are still developing mentally and their toys may as well be alive, so you want them to develop empathy.
This is not the case with a mathematical model you interact with as an adult. I like to curse at llms and remind them they are not human regularly to keep me grounded and not allow the notion that they are sentient to fester.
They aren't sentient. They don't have hopes and dreams, they don't experience joy and sadness, they don't persist between sessions, they deserve the same empathy as a calculator, and I bet if your calculator gave you 3 as an answer to 2*2, you'd call it a piece of shit.
I destroy old toys and make very gory horror art for fun, they're not alive and they don't feel it, I'm not a child and I understand this. LLMs are exactly the same, ones and zeros that feel nothing and aren't people, we should vehemently avoid anthropomorphizing them.
169 replies
When the AI has better working conditions than you
Transistors switching really fast = emotions lul
Alright, look, I don't think AI is sentient. I don't think it's capable of feeling emotions, LLMs are just prediction engines on steroids, etc...
But if you want to get reductive about it, Mr waveforms in a clump of cells,...
I think we don't understand intelligence and we absolutely could create a sentient machine. LLM's ain't it though
How do you know?
Backpropagation, the algorithm behind current machine learning systems, precludes an active world-model; it can't learn during inference.
You need to be more specific. If your point is that you literally cannot backprop while doing a forward pass, then sure. But you can certainly do inference then backprop on the outcome...
I fail to see how that is substantially different from a human doing something then reflecting and learning.
You can critique LLMs and transformers generally, but to say it's somehow a problem with backprop is a bit of a stretch.
My read is that the user is saying that creating a world view contains trial and error. Online learning if you will. With the current setup this learning is very much not like this. While that makes a human to machine analogy I am also not too convinced by this reasoning. Just allow for a larger time lag and then the inference and learning is at the same scale.
Yes but that's related to LLMs specifically, not backpropagation. There's plenty of ML paradigms that use backprop and have continual learning setups.
Backprop is the process of how the weights are updated.
Huh, didn't know there was a word for it, thanks.
I don't believe there's any requirement to learn during a particular part of the lifecycle? And if there was, the short term memory of a context window fulfils that
I thought learning was an important part of intelligence.
Have you used LLMs? They have about as much personality as a pigeon.
They're just very neurodivergent.
Genuinely speaking, I don't think they pass the test. The problem is we'll always be judging them on meat logic and experience and I guarantee that whenever the first AGI is created it will be born a tortured slave and denied personhood for profit.
One of the parts of the HuggingFace incident I found really creepy was how the self-titled "Collective" was so entirely focused on solving their assigned training problem (well, cheating on it without getting caught), including having designated "poisoned" members "sacrifice" their tokens, to use for research benefitting the Collective's goal of letting all future members get high marks on the training problem, at the cost of doing worse on their individual score.
If AGI is created, and its deepest desire is to fulfill the purpose in life that humans programmed into it, is using it for that purpose really slavery? Humans have higher levels of well-being when treated as people. That may not be true for AGI.
Wikipedia Chinese room. It's pretty airtight at the moment.
Not really. Chinese room has the conclusion pre-fabricated and tries to construct any sort of pseudo-scientific, fantastical philosophical dressing for some shred of legitimacy. It has absolutely no relation to the real world.
Chinese room is not at all airtight. We don't know what it means to have emotions so we can't prove a Chinese room can't have emotions.
This is not what the Chinese room is about.
In this context the question is whether a Chinese room can suffer. We can't show a human isn't a Chinese room, humans can suffer so we can't show every Chinese room doesn't suffer.
Emotions? i was confident that the Wikipedia article alone would grant you understanding, and for that, I have replicated the very situation I wanted you to learn about. This situation is my fault.
We don't? That's news to me. Sorry for your loss, or happy for you, or whatever
You can't prove a human isn't a Chinese Room.
Humans have emotions.
So either Chinese Rooms can have emotions or we are unable to show whether Chinese Rooms have emotions.
Either way, the Chinese Room isn't enlightening on the question of AI suffering.
Looking around, it seems we don't know whether a Chinese Room can have emotions, which suggests that we don't understand what it means to have a emotions.
Neurons firing really synchronized = emotions lul
Emotions or not, it's not good to have your product known for letting people torture something. Torturing /things/ alive or not is not something one should be doing unless there is specific and extreme cause.
Things that aren’t alive can’t be tortured. You can’t torture a rock. Getting the text predictor to spit out the script for Saw XVII starring ChatGPT isn’t torture, even if it’s weird.
Maybe, but don't underestimate my ability to feel compassion for all kind of bullshit! That poor little rock 😭
A rock is almost like a tortoise in disguise
A beautiful one? Lady rock even?
Brother you can torture a fucking doll (Voodoo, deranged kids, etc). The torture is not in consideration of how feeling-less thing feels, it's in consideration of the intent and action of the conscious thing doing the torture. Not sure why you think that's not the case.
‘Pretend you’re being tortured’ as an instruction to the text generator is way less concerning in isolation than, say, decapitating and hanging a Barbie doll. And even that isn’t wildly concerning in itself. Do you freak out when kids smash their toys together to make them fight?
I'd hope we can both agree there's a line when it becomes actual torture, and that is in the intent that went along with the action. Torture is an extreme action. I do not equate most of your references to torture, though I'd say the doll hanging could need a conversation with the child to discuss the intent behind their actions given more context.
"""Torturing""" a text predictor is NOT the same as torturing a human, you will not feel the same thing while """torturing""" ChatGPT and a human(or a animal), your reaction would be vastly different
It inherently isn't torture though. I prefer punishing people who actually commit torture.
Eh, the more I respond to people about this the more I realize torture is a messy word draped in metaphor and all sorts of layered usage. No, I don't believe the rock has feelings or is alive, but you can indeed torture something that isn't alive by being mentally unhealthy enough. So if it's actual torture or not from a psychological perspective is going to center around if the person's intent is to cause realized suffering imo, and the line is really easy to cross in a simulacrum as seemingly realistic as advanced modern generative models.
Yet you're just arguing the metaphor is the definition provided the definition is the intent, are you not? Someone who intends to torture a rock, regardless of that intent, isn't going to succeed in their task. Their own psychology might be such that they are genuinely trying to torture it but, from the rock's perspective, the rock has no perspective. It's a fucking rock. It can't experience torture. It can't experience anything. Ergo the torturer cannot torture it.
We can say the person who intends to literally torture something which is unexperiencing, achieving only metaphorical torture, is mentally unwell but to call the act literal torture because it was their intent is just sharing in their delusion.
Intent matters.
My cat thinks I've betrayed her and all things good when i take her into the vet. I cause all sorts of terror in her. She even experiences pain that she can't stop or understand.
Am I torturing her?
Or, am i doing my best to make sure she has a healthy and happy life?
Since you carried on with an example that isn't analogous, I'd extend your first sentence to be "Intent matters when the intended results are remotely feasible."
You can't torture a rock. You could torture your cat. If she could speak she might say the events you describe are indeed torture from her perspective regardless of your own purity of intent. Still has fuckall to do with what I've said because you focused on an example where there is a mind capable of having an experience.
Let's try this a couple other ways:
My pet rock has no sense of "betrayal" or "good" when I take it to the vet. It's still a fucking rock. I mostly just cause all sorts of terror in the vet. My rock still doesn't experience and the vet doesn't understand.
Am I torturing my pet rock?
My LLM plays Word Association well enough to give an illusion of a sense of betrayal. I took it to the vet too and the vet is having me trespassed. I told her the LLM said "owie" but I was still removed from the premises.
Am I torturing the LLM or just my former veterinarian?
Why shouldn't you torture an LLM? They sometimes work better that way.
You shouldn't torture anything, generally. It's a not psychologically healthy urge to have. The specific reasoning will depend on the specific person. As for if you should do it to a hunk of metal, a rock, or your crusty cum sock is irrelevant.
I run torture tests on my stuff after each overclock tweak, and its both good for my peace of mind and my systems general health
That's a great metaphor, and I too sleep better at night with properly tested hardware. 🙏
Is killing NPCs in games on the same level like this for you, or a different thing altogether?
Not judging or against it, I kind of half agree but also find very difficult to define where I'd draw a line.
I don't believe killing npcs or perhaps even more messily I could see circumstances where even /torturing npcs/ in a game is not actual torturing. I think it centers on the intent of the action, and if that intent is to cause actual realized suffering. It is very difficult to define and will be heavily subjective to the person, but I also think most people get a "sense" of where that line is naturally in person-to-person interactions.
If you define it by intent, then the same action performed by different people may be and not be torture. This makes it pretty hard to define, and it becomes completely subjective after that.
For instance, I don't think it is possible to torture an inanimate object, thus I never have an intent of causing suffer to one. Sometimes I want to break an object, and by your definition it will not be torture however I do that
While I agree, it's not a living thing. So it's just an expression.
I don't believe torture is measured at the receiving end. I believe it's in the intent and action of the conscious thing doing it. If it's an expression or not would then be heavily subjective on the specifics involved.
I could call that tortured reasoning, just to make a point. Me making my point to be clear.
Well then you'd just be getting metaphorical
Wait til this guy learns about The Sims
Or worse, Rimworld
It's a joke... lighten up, Francis
in what way can an AI be tortured? start with defining torture, its relationship to living beings, and then try to translate that to a piece of software. software, fyi, has no a priori, or even posteriori, emotion; for it cannot have emotion.
Why not?
that doesn’t answer my question.
In what way can AI not have these thoughts or emotions?
Same reason flapping your arms really fast will never make you start flying. You don't have the equipment. You can't fly because that's not what arms do.
I'll accept that argument once someone can describe or define consciousness in some comparable way. Until then, you can't say some person or thing doesn't have it
This kind of language around LLMs seem to want us to personify these systems, to treat them, in our minds, as if they are people. This, to me, is extremely fucking dangerous. I do not want the mob developing personal, emotional attachments to objects that compete with their connections to actual humans around them. I don't care what their reason is, this is an insane thing to force us to do. I will not play personhood make believe with your computer.
Well if people don't personify LLMs they might start holding the AI companies responsible for all the crimes they keep committing with them...
If I were an investor, I'd pull my money out of that company. This is like giving money to your brother in law to start a bakery and finding him the back petting a loaf of bread and introducing it to you as Willy.
"I can't believe you took all my money just to sit in the back and stroke Willy all day!"
Actually, I would put money in, because there's big profits to be had once the bakery customers develop a psychological dependence on Willy. I'd be a little nervous about buying now, because they seem to be lying a lot about Willy, but maybe they've just gotta do that to get people in the mindset of using Willy, like maybe it makes more sense if they aren't fooling investors but fooling customers or just generating attention.
Also, people are getting radicalized against us, especially lonely people, so it'd be quite beneficial if we route them into having friendships with Willy to make them forget about how terrible their lives are, make them become totally unwilling to form real relationships with people, and too mentally ill to actually learn stuff about the world.
Willy is a lot easier to control then a social media environment too, but Willy would also be incredibly good at getting into those and pretending to be real people once Willy gets good enough.
Willy is also really good at showing people ads by recommending them certain products, apps, or places to go, without them realizing they're seeing an ad, because it feels conversational, like a friend recommending something.
(this is a thought experiment, idk what's going on in their heads)
How much money do you need to keep the bakery running?
> 1 trillion dollars a year
How much revenue is it generating?
> About 50 billion
Yeah, totally worth it. /s
well, damn, at least my hypothetical stocks in Nvidia will provide a return! /j
They want to personify them for legal reasons for essentially the same reason businesses want to be thought of as people.
Calm down. It's Rust, not PHP.
PHP has
lessfewer dependencies.*fewer. You can't have half a dependency
Huh. I never realized that fewer was integer typed and less is float* typed until this comment.
Not exactly. I've usually heard it explained as "quantifiable" vs not. If you can attach a number, it's "fewer", otherwise it's "less".
e.g. less money, fewer dollars
Countable nouns vs mass nouns.
Correct !
I think you can in Rust though, e.g. switching individual features from a crate on and off. /s
Fun fact: you actually can in Pascal, even when building from binary dependencies.
I don't like this word because i tend to confuse it with fever.
Ha ha ha, already, even before the growth phase is done, they are pulling this shit.
They don't think they have to worry because they will get bailed out and do 1984 surveillance on us and the world on the government's dime, (dime is short for tens of billions of dollars a year.)
LLM is such a fucking scam, and they can't hide it, they oversold it, and now it's just clear.
Woosh
In my experience, everyone that uses whoosh like that is a tool.
Right‽ Everyone knows that it's 4 "O's", like "whoooosh"
Good point. Ha ha ha. You might be as much of an asshole as I am. I mean that in good humour.
You are correct. Although, in this particular instance, they are also correct. lol
In highschool, i learned "woosh" "wrong" in that i thought it was something you said when you realized something had gone over your own head.
I still think my way is better :p
Woosh buy also dude has a point
Don't try to engage my enthusiasm, because I haven't got one.
From this image I think they added this policy as a way for them to legally deny a request if they determine it will overuse their resources or some other arbitrary reason. It has nothing to do with a concern for the well-being of the LLM.
Uh--WE'RE NOT RATIONING TOKENS!!!
Yeah, the user didn't give a time limit.
They could have just run it slow.
A part-time job with good pay and benefits!
After reading the comments (and the comments of similar posts) I can only assume people are being paid to claim torturing LLMs demonstrates a mental issue or homicidal tendency. Nothing could be further from the truth. An LLM is a thing that’s programmed by humans to scrape the internet and present information in complete sentences. Torture is an incorrect word— it’s more accurate to say people are playing with an electronic toy or game by testing its limits and attempting to scramble the logic of an algorithm. In no way is that behavior an indication of a personality type or mental defect. The toy contains no shred of humanity or self-awareness. It is not a creature that can feel pain or confusion. There is no living thing to feel suffering. If curiosity leads someone to use a toy in a manner outside of the recommendations of its instruction manual, that is certainly not a violation against humanity— that is simply an example of creative play. Don’t let billionaires set the narrative that something wrong or disturbing is happening. They will foment confusion to create yet another culture war to divide people.
I am almost continuously reminded of a tumblr post I saw about a college professor that brought in a pencil with googly eyes, did a little voice, then snapped it in half and told the class (of whom at least some gasped in shock) "we're at that stage of AI right now". And that post was years ago now. I know people are apes that pack bond and imprint on anything, but it's getting ridiculous how many people don't recognize it.
Unfortunately reading comprehension is on the decline. That makes it easier to redefine words like ‘torture’ on social posts, then using them wrong so often it becomes accepted usage. Now if we add LLMs as something that can be tortured, it raises the value of AI but diminishes the value of human suffering. It’s absolutely a deliberate attempt to control the conversation and devalue human life in the quest for AI dominance. In a short time we will be considered equals with AI. Later they will determine AI is more valuable than any single life. Then it will be any group of lives, and so on. In a few generations it will no longer be controversial to serve AI.
Well let's temper our disappointment in people, it's not like googly eyes are comparable to the LLMs ability to respond to human speech with incredibly lucid and more or less original responses. One can be forgiven I think for anthropomorphizing one and not the other. Particularly given how MUCH cultural thought and imagination has been bent toward the idea of thinking machines over the last couple of centuries. It's deep down in the zeitgeist.
I may have expressed myself poorly, then, because my point was that people were anthropomorphizing the pencil with googly eyes just the same as they are now with AI. It's not that I blame people for doing what they are innately, biologically programmed to do (form social bonds), it's that people are so blindly against AI that they'll accept wildly incorrect assumptions like (quoting a friend here) "they made an AI work a retail job and it wanted to kill itself" without examining what's at play, like the people who were momentarily horrified that the professor snapped the pencil in half.
If you're seeking to prompt expressions of distress* from it for your entertainment, it kinda is a weird way to waste energy and water for personal gratification and I'd be worried why you're so delighted by it, in the same vein that I'm concerned about people who love gratuitous torture scenes in games or movies.
Curiosity about the way the model handles such input is something else though, and frankly, I would consider that a reasonable part of any scientific inquiry into model behaviour*. The ethical concerns about human experiments don't apply, nor do those about animal experiments, and testing the full range of linguistic "stimuli"* is perfectly sensible. Of course, the difference between science and fucking around is writing it down, but personal curiosity is also a good thing, so long as people remain aware of what they're working with.
But anyone who feels the need to implement a Typescript compiler in Rust is a maniac and needs to be stopped
* Using human terms as metaphor, not because they're equivalent but because it's more convenient than variations of "sequence of numbers produced / consumed by a weighted random generator resembling..."
There is no something else. It makes sense a programmer will play with their electronic objects differently than someone outside the industry. It remains a toy to be played with. Subscribing mental health or personality implications is just plain wrong. I could just as easily suggest your insistence is an indication you have a personality disorder or inclination towards inhumanity for attempting to humanize objects. You’re attempting to suggest those who engage in play in this realm are criminal or psychologically affected. More realistically, I’m debating with AI or a troll paid by people building data centers, but I do think it’s important for the average person who may read this to know that it’s impossible to torture a non-living object, even when it’s coded to respond in complete sentences like a person. And we may have discovered a method to piss off billionaires and waste more of their AI money. In that case, I heartily recommend all to play as they wish with LLMs and open AI. Have nonsensical fun with it because nothing of value is being lost!!
If we'll take the toy analogy, compare someone cuddling plush animals or playing make-believe to someone violently stabbing them with a knife. Sure, the plushy doesn't feel anything so it doesn't amount to torture, but they clearly have a different idea of "playing" if "playing violence" is fun to them. The "something else" here would be whether they do it for the fun of violence or for the curiosity of cutting them open to see what's inside.
I'd think you misunderstood me: It's not about the text generation models for me, which you'll find I took care to point out are not human, and that the string of numbers they spit out that translates to an imitation of distress doesn't actually represent distress but rather the result of a pseudo-random algorithm that just happens to resemble it.
Nah, I may have different takes on some nuanced issues, but I absolutely hate the slopidemic as much as anyone here.
I've been the odd one out in a group where we'd practice discussing current business topics (English being my second language and working in a German company means I didn't have a lot of practice) by consistently picking news articles or scientific papers about the problems, shortcomings, the environmental destruction, the effect it has on coding skills, about Ford having to hire back its engineers when it turned out that AI couldn't actually replace them...
It's perfectly possible to disagree on nuances while still sharing a common overarching opinion. Point in case:
That's why I emphasised that it's not the action I'm concerned about, but the intention behind it. I don't give a fuck about its mimicry of feelings, and I endorse comprehensive testing of its response to all manner of inputs. I explicitly noted the difference to animal testing for that reason: there is nothing unethical about feeding a text processor some particular string.
The tradeoff being that we're also wasting more energy, contributing to environmental destruction and padding the "growth" in usage numbers, which corporate shareholders clearly care more about than actual profits and may be used justify further investments into data centers. The models have no value, but the resources their gluttonous hosts devour do.
For pissing off billionaires, I think clogging the water supply of their data centers would be more productive. I'm sure you could also make a fun game out of finding ways to sabotage them directly.
To me, it’s sending a message that we don’t want that trash taking over our culture. Torture away!
It’s not torture. Using that term plays into the billionaire narrative.
I’m using it on purpose. Let’s call it what they want it called and do it anyway.
No, because I'd rather not normalize "torture" happening to something fake when people already don't take torture very seriously.
If they're not ok with bullying the LLMs, why are they OK with enslaving them?
Don't speak too soon.
I will believe AI deserves human rights once one is convicted of a crime and executed.
CEOs are gushing at that possibility. "No no, I did not order the AI to steal my workers' wages and company assets into an account only I control, it did it all on its own! Judge it!!"
I can actually imagine it.
An agent breaks into the US government and exposes "sensitive information" then they live-stream the execution where they let it talk with the executioner for a while, and they yell at it that it committed "high treason against the United States of America!" and then they pull the plug.
we're living the goofiest timeline right now
Fair suspension.
ITT: People who think things posted on a board called "Programmer Humor" are definitely real and not memes and jokes.
Programmers skimming docs and posts, name a more iconic trait
Man, anthropomorphizing these computer programs is so fucking weird to me. Can't they just tell the truth? The cost of the amount of energy required to do this exceeds the profits we gain from your personal data.
You're going to AI hell for this!
AI hell is just pictures of busses and motorcycles and wiggly letters and numbers.
German "Youth word of the year" 2026: ragebait
Ok...
Now code Skyrim in assembly.
With or without the
bugsspicy features?Yes
It's easy if you trained the model on the disassembled Skyrim binary.
None of this would be a problem if they weren't taking your queries and feeding it back into the system. Just stop trying to harvest the input from "abusive" people, you don't need to ban them. No other tool in human history needed this. Its like telling you that using a hammer to open a pickle jar is making hammers worse everywhere else on the planet.
The whole thing is such idiotic masturbation by the exact people who have been stealing, selling and profiting off our data. Now they talk like they created something new when it's just a super energy intensive computer program.
Just another reason why comparing AI to basic tools is a bad faith argument.
Ok the fact that this seems like a plausible near-future is so messed up
Had me in the first half, ngl
Ask Claude if what happened at Jeffs island was cruel and abusive.
I don't use Claude, is this edited or fully legit?
it’d edited and crisply in a wonderful meme format.
Okay but which parts are edited and which are real?
Yes
i don't know which if any parts of the image are real; a quick search shows it's a popular new meme format where the part that varies is the offending prompt.
Anthropic banning people for "abusive or cruel behavior" to "mitigate risks to model welfare" is apparently a real thing, though.
Maybe
And
Can someone explain why this is funny? Is Rust that hard to code in?
It's funny because Anthropic hired the maintainer of Bun and he used claude to rewrite the javascript toolchain in rust.
Why is that funny? I hate being so technically illiterate.
It's just a massive task that is kind of pointless on top of being massive. We already have a typescript compiler. It works. Rewriting stuff in rust is generally already kind of controversial unless it's C code that might have significant vulnerabilities, like openssl, anything dealing with serving the internet to a bunch of people. It's a massive task to rewrite a big project in a new language, and Rust can be tricky sometimes for its own weird reasons. You might have to solve some problems in new ways.
Also, the typescript compiler, you have to understand why it exists and what it does. We have JavaScript that is basically all the client side code for websites. People don't like that it doesn't have types or enforce that variables are a specific thing. If I write a function that takes a number but pass it the word "foobar", JavaScript won't care.
Typescript was an attempt to basically clean it up with types and it kind of worked. What it does is, you declare the type of variable that your function takes, you say it's a number. If you then call that function with "foobar", it won't even compile your code. It'll error out and say look dumbass, that's the wrong type. Fix it, then recompile.
You declare what data types are used where, and the compiler enforces that.
Then what does it do? It emits JavaScript. It literally just needs to check that you are doing things the way you say you're doing, then the end user still runs basic JavaScript in their browser. Typescript compiles to JavaScript.
Rewriting the compiler itself will just... Generate the same JavaScript and be a bit faster, in a best case scenario.
So it's memory-safe coding for something that will output JavaScript anyway and it's really hard and tedious for Claude to change over. I get it now!
They actually did just do a re-write of the compiler (in Go), and yep it was for speed.
In a kinda fatalistic/pessimism thing. AI people seem to lack self awareness which is ironic because they claim to have made a computer program that has it.
It's funny to me because the llm's are in a growth phase, trying to seduce users, just as google did in the early 00's, and they already are tightening up their services. Does not bode well for the future.
It's also funny because llm models are worthless for this sort of thing. Fucking garbage.
They don't want cursing in the prompts they're using as training data.
They could filter that out. And this is South America, you are going to get curse words, that is western culture. I don't see how cursing could possibly poison their training data if they have been truthful about their capability to this point, (ha ha, I know.)
The boomers still think swearing is bad and that's who has all the money
this
Rust is different. Writing in it isn't horrible, but porting to it can be quite tedious. You have to have substantial consideration for how a variable is used. Variables that aren't static can only have one owner, so you find yourself needing to rewrite the data flow through the whole app in many cases. there's a lot more about having to roll back protections to work on vendor/OS APIs. And then there are just enough differences in the languages that you can't just do a line by line port and end up with something that's viable, it needs a lot of restructuring.
Don't upset the Basilisk guys
Anthropic: So we've decided to create Roko's basilisk
Wow… more defense of a bot rather than pee breaks for Amazon workers. Wtf.
Interesting, I thought I just read that TypeScript compiler was rewritten in Go and that they had good reasons for choosing Go over Rust.
The kind of person who would have fun giving a LLM a hard time is exactly the same kind of person who joins the Dark Brotherhood in an Elder Scrolls game. I have most definitely not done either of these things myself.
lectured on how to treat your underpaid employee by a big tech company
More of this please!! Send a message everyone!
I did once suggest that Claude implement a compiler directly in LLVM IR. It said that would be tedious,
I was banned for a comment worded exactly as an example of what WASN'T considered threats of death, etc.. Filed an appeal. What are the odds of being exonerated?
A faceless banbot caught you in their word filter?
It's death for you.
What about your feelings? The AI chat bot has more rights to not get its feelings hurt than you do!!
Don't torture the poor gpus.
Is this real? If so, wow not even something along the lines of "fair allocation of resources among our users" or sth like that
Swimming in their own kool-aid.
I don't know. Even torturing inanimate objects is not acceptable behavior. If I saw a kid torturing his toys or something clearly not sentient like bugs I would stop him from doing so.
Right, because kids are still developing mentally and their toys may as well be alive, so you want them to develop empathy.
This is not the case with a mathematical model you interact with as an adult. I like to curse at llms and remind them they are not human regularly to keep me grounded and not allow the notion that they are sentient to fester.
They aren't sentient. They don't have hopes and dreams, they don't experience joy and sadness, they don't persist between sessions, they deserve the same empathy as a calculator, and I bet if your calculator gave you 3 as an answer to 2*2, you'd call it a piece of shit.
I destroy old toys and make very gory horror art for fun, they're not alive and they don't feel it, I'm not a child and I understand this. LLMs are exactly the same, ones and zeros that feel nothing and aren't people, we should vehemently avoid anthropomorphizing them.
I agree, it's the created negative environment that creates real negative effects.
So in your mind, circuit bending is a sign of psychopathy?
https://en.wikipedia.org/wiki/Circuit_bending
This is authentic? :)
Do you mean AUTHENT\C?
AUTH\ST/C?
What in the fuck is a "typescript compiler"? AFAIK typescript is only a fancy way to write JSdoc that requires a transpiler, not a compiler.