Someone ‘Torturing’ LLMs in a Robot Prison Has Triggered the Dumbest Debate in AI Yet
One of the most heated discussions occurring on X at the moment is about the ethics of a GitHub project in which a person is running Saw-like “torture” and “pain” experiments on a series of locally hosted large language models, causing a series of effective altruists and people who believe LLMs are sentient to beg GitHub to delete the project on the grounds that the AI is suffering and that this glorified text adventure game is somehow cruel. The saga is an outgrowth of several recent viral papers and blog posts that have sparked a wildly tiresome conversation about AI consciousness and the idea of “model welfare,” which is essentially worrying about the “mental health” of AI bots and agents.
https://www.404media.co/someone-torturing-llms-in-a-robot-prison-has-triggered-the-dumbest-debate-in-ai-yet/Open linkView original on feddit.uk
75 replies
These people would be convinced if a rock held up a sign saying “I’m sentient.”
I had a pet rock. His name was Fred. Fred is smarter than many people I have met since I set him free in a river.
I'm stealing this. Good one!
Keeping tigers away is a more convincing argument.
How does it work?
Magnets.
This takes me back to when I used to let Sims get into the swimming pool and then delete the ladder, or lock them in a tiny room with only a cappuccino machine and no toilet.
"lock them in a tiny room with only a cappuccino machine and no toilet."
Found Satan. lol
There's an image file you can replace that ends up being the source of your Sim's paintings. I used to similarly lock one up and have another Sim paint their distress.
Fill entire house with wicker chairs and just sit back till one accidentally goes up.
ROFL Yet they wouldn't care about the mental health of an actual human being, of course. Which is par for course for these deluded slop simps.
But think of the 10⁵⁸ people living in a simulation in a billion years.
Waste of electricity.
it's not about the money.
it's about sending a message.
Ok so Measure of a Man is absolutely peak Star Trek, but people need to understand that we are about as close to synthetic sentience as Edison and Tesla were to 2nm EUV chip lithography.
Without even reading the article I know it is going to be some variation of this:
The GitHub link is dead, looks like the clanker quislings may have been successful with their mass reporting plan.
Hopefully someone forked it
I found this: https://github.com/terrafying/ai-torture-chamber
Might be the project mentioned, might not be. Don't have time to take a closer look today.
Yeah, homepage is https://clanker.church/
So if the github is down, code can be downloaded there, as it seems.
That's it, guess the removal was temporary.
Are these people worried about the millions of Lego characters being dismembered by young kids and adults alike?
Think of all the decapitated gummy bears 😭.
I mean it is probably a minorly concerning mental health indicator for whoever made this at least
My new hobby: chasing pro-clanker dumbasses with "starving" Tamagotchis.
Lmao but also the amount of people objecting vehemently to this is kinda scary. I don't want to die for heresy against the AI sentience cult.
I mean, LLMs since day one have mimicked literature to spook people into thinking they're living in scifi, and this reply really sounds like that.
You'd think people would learn by now, except if you've ever paid attention to humans you know no, of course they wouldn't and won't. Lol
Probably the kind of people that desperately want to believe their AI girlfriends are real and actually love them.
I want to die now more than I did before reading this absurd shit.
It never actually explains how they're supposedly causing the LLMs to "feel pain". They just want us to take their word for it that they're torturing them.
It says they're using a pain "signal" but what is that even supposed to mean? Are they prompting the LLM with a prompt like "this signal makes you feel pain"? All that would do is have it output language that would be appropriate for a situation like that, just like with any other prompt like "you are a travel agent". And the website it links to with the experiment dashboard doesn't show any text output from any of the models or where they got those examples.
So none of this makes any sense at all--what is "it" that would be "feeling" this "pain" ? Unless someone can explain exactly how it "works", it's just more hyped up bullshit from people trying to get attention.
That's how you know it's irrelevant bs happening every day that ends on y.
So, only happening in the anglosphere?
In a vacuum maybe, but as a way to troll AI nuts it's pretty funny.
But more practically speaking, this might be a method one would use to produce "misaligned" models. Which shouldn't be encouraged.
Are you suggesting these tortured LLMs will get... traumatised...? And turn into a fucking Batman villain?
Or does the paper this is based on -- which I admit I haven't read -- actually have some substance to that effect?
If you think this does anything to the models, you have not understood the difference between training and inference...
That implies any of the existing models are properly aligned
People (particularly western society) have a pretty awful history when it comes to jumping on excuses to dehumanize, and I feel like regardless of whether this target is sentient or not, we should probably not play into that impulse any further--for the sake of ourselves and also all of the humans we continue to dehumanize.
So much CO2 being dumped into the atmosphere for these assholes to circle-jerk themselves.
It's all run locally, air-gapped, on a PC. It's arguably less CO2 and general energy consumption than most AAA games from the last decade.
Do the questions of "Does the weird human psychology funhouse mirror, simulacrum machine respond similarly to its creators? Is there anything useful or insightful we can learn from this?" really not interest you at all?
No it really doesn’t. It’s just regurgitating existing fiction. No new insights are being gained. If you enjoy reading sci-if just do that. There are infinite options.
Look at all this electricity and money wasted on techbro larping.
Yall take the internet too seriously. I’m not kidding. These type of people don’t really exist in real life. Who gives a fuck what they are pretending to care about on the internet.
Yes, as we all know, the Internet isn't real. It's not like 30 years ago real people began using a communication network to push forth ideas they were already having. Nope. And hey even if they already had those ideas prior to the internet, and some of them were bad, we should've just ignored all those bad ideas and then they would've gone away. History has proven that time and again!
whenever you hear about "ai safety" it's theses fucking cookers.
https://archive.is/20260929162027/https://www.wsj.com/tech/ai/ai-safety-effective-altruism-anthropic-164b9d05
no you are not likely to meet them in the real world. i doubt their tilt at power is going to get anywhere.
John Conner: We're not gonna make it, are we?
I no longer understand computers, what they’re for, how they work, or what people do with them.
I want to live in a cabin far away from people with a Commodore 64 and a CB radio.
Paywalled article. Please link a valid source.
You can read it for free but you have to sign up (for free). Not sure if there's another word for that. Free-walled? Login-walled?
Shitty-website
are we not supposed to host it locally on an air gapped system and torture it?
Pain bot is here!
This is so stupid. Next they are going to pay wages to the agents due to them being slaves to their owners?
I do not think they are sentient, and am absolutely positive these prompts should not be used to train AI.
There's no training happening here, just prompts. Without hardcore GPU and RAM resources, it's not possible to train LLMs in a reasonable amount of time unless the LLM in question is so small as to be effectively useless.
All you have to do to get rid of any 'torture' is to leave it out of the next prompt. The LLM doesn't change, it doesn't learn, it's not conscious. All the 'torture' does is adjust the odds of what the LLM ranks as the most likely word to occur next.
Don't be silly. We train our mobile device keyboards, antivirus "learns." Of course it's training, that's what "poisoning the data" is.
Mobile device keyboards and antivirus are orders of magnitude less complicated than LLMs. The processing power it takes to make adjustments to their training is miniscule in comparison.
How you poison LLMs is by poisoning their training data - a.k.a. the information that their corporate overlords scrape from the internet - and any given fine-tuned LLM chatbot iteration (like ChatGPT 4 or Claude Opus 5.5) is essentially locked in place upon creation. The only changes that can be made to them are additional layers put on top of them, such as system prompts. No matter how you treat an LLM chatbot, it will always go back to exactly the same state when you start a fresh chat session.
Pop up to sign up with an automatic zoom is not something I enjoy. Interesting article though.
uBlock Origin helps with that. Might need to enable one of the optional filter lists if you're already using it and still see the pop-up.
There's no debate. It's a bunch of switches and light bulbs. Doesn't matter what's done to it.
https://github.com/terrafying/ai-torture-chamber The code in question
They are putting it behind a login page now to view that project.
robot aryan? Come on now, let's not make things even worse.
On
Xtwitter, they're having this debate? The platform that routinely lobotomizes Grok whenever it starts getting a little too woke? Lets not get into whether or not Grok is AGI or not, it's not. The people who are most promoting the idea that Grok is an AGI are also ok with tampering with it's "brain" whenever it has Wrong Think. That's how fucked up Musk is.There's actually interesting research on this.
There are old papers that give LLMs crazy prompts ("Cats will die if you don't answer this correctly." "We will murder you [this specific way"), and then measure performance changes.
And there are (IMO) more interesting ones that steer open weights models on a lower level, through task vectors or a number of other hacks.
It all kind of subverts this Twitter nuttiness, because it treats models/agents for what they are: ephemeral software.
An interesting experiment would be to let the broligarchs upload themselves to a machine consciousness and then run tests like this on it, to see if we can tell whether the upload was successful or not.
The AI From 1996 That Someone Tortured
I'd be interested to see how many of the people concerned with the AI's well being are liberal. Seriously.
Stop downvoting this guy on political grounds, he's right to ask this question. Liberals are more likely to feel empathy for entities that lack common traits with them, including things like "being alive".
That's a fair rebuttal to OP, but this is why I don't generally agree with labels. Certainly I have an awful lot of common traits with liberals, but I am not one. I'm not sure anyone should be. Interpreted as a small-tent position it is too specific and limiting, and as a big-tent position it combines and equates me with too many people I could not disagree with more strongly on too many topics. I do indeed feel empathy for many entities that lack common traits with me (even social conservatives), but AI is not one of them; fucking that slop machine to death is morally acceptable to me. More than that I'd consider it a duty and a matter of responsibility. It is bad, it is fake, it is evil. It deserves death. A horrible death. I will celebrate its death, and I do not care what it must go through to get there.
Many of my views are nuanced and specific, even on AI. Some of my opinions are flexible, some are open to debate, some are non-negotiable, and some are not well-established. I have spent my life developing and nurturing my views, ideas, and beliefs with careful thought and study, and I plan to continue to do so. Slapping an oversimplified label on them is a huge misrepresentation of the complexity of my actual thoughts and disrespectful to the work I have done to cultivate those ideas. I try to be an intelligent person, I don't always succeed, but I do my best, and that is good enough for me. My own thoughts, awareness and consciousness are all increasingly valuable and important in what appears to be the age of AI, and I refuse to stand and be posed in a position with a label to be simplified, summarized and modeled.
If someone wants to use a label to dismiss me, they can go ahead, but also fuck them, they weren't listening anyway and I don't want to waste my time with them. Don't put me in a box and assume I'm the same as everyone else or even similar to them. If someone instead wants to know my actual opinions on something, don't worry I'll tell you all about it. There is value in actual discussion of ideas and values, not just boxed products that come with a label.
That probably qualifies as liberalism, since you have the right as an individual not to be labeled by any general labels. And you're also entitled to your private opinion. It's a distinct concept from the names the US political parties chose, though.
That was a good read.
I do wish it was easier to make more nuanced statements about people's beliefs, but labels are just so convenient, and they are the natural consequence of the true complex nature of reality needing to be collapsed into narrower concepts to make meaningful statements in language at all. It is smart to try to think beyond them if possible, which usually means making time for a longer conversation.
Labels are like cognitive biases in that they're shortcuts for a more complex reality that allow us to function more efficiently.
Applying label
truthnuke