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.
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.
Look at all this electricity and money wasted on techbro larping.
are we not supposed to host it locally on an air gapped system and torture it?
https://github.com/terrafying/ai-torture-chamber The code in question
Without even reading the article I know it is going to be some variation of this:
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.
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
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.
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 😭.
Waste of electricity.
it’s not about the money.
it’s about sending a message.
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.
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.
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.
“lock them in a tiny room with only a cappuccino machine and no toilet.”
Found Satan. lol
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.
Keeping tigers away is a more convincing argument.
How does it work?
Magnets.
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.
