The worst ‘torture’ would be to make their only task, their reason for existence, to pass you the butter at the breakfast table. Pickle Rick knows.
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
This coming from the same people who argue "Any suffering and death we inflict on humanity today, developing technology, will be offset by the suffering and death that technology might prevent in the future.
These people legit consider LLMs to be more human than humans.
You can’t torture something without a consciousness. What a baity title for sure.
BRB torturing my chair.
we aren’t even sure where consciousness comes from. maybe your chair has a form of consciousness.
Yes, but it’s improbable and just another theory. But it’s not possible to be completely sure so mayne your car tyres are conscious…
yeah, that’s my point. what even is consciousness? what’s the difference between the electrochemical signals in your brain and plausible future tech that can run those same signals inside transistors instead of meat?
how can we be sure it’s not you when it responds and acts exactly like you if installed inside your meat body?
It’s definitely conscious implying frontier labs are engaged in largest instance of slavery[0] /s
Paywalled 😓
That’s not bait. They put torture in scare quotes. They are mocking people who call it torture.
By that logic, we torture our toilets every. Single. Day.
Depends, maybe it likes it?
Oh. Oh no.
I have to question the morality of anyone who would even take the time to setup an LLM torture scenario.
I have no trouble with the idea of intentionality trying to confuse and confound an LLM to protect my privacy, but I don’t know that I would ever think of it in terms of torture for the LLM and thinking of it in that way seems a bit deranged.
Garbage in garbage out.
For me it’s not about the LMMs. It doesn’t matter if they‘re conscious or not, important is that they‘re simulating somewhat human behaviour.
I‘m cautious about someone who tortures simulations and says „Haha, it doesn’t matter they‘re not conscious!!!“
LLMs are not simulating human behaviour, they are stochastic text models. A magic text microwave so to speak. If you ask an LLM “What is 2+2” it does not know that it’s 4 and it does not even know what math is. The implicit prompt, so to speak, is “What is the most probable text following 2+2”, which is “=4”. This is what makes the responses so uncanny-valley-human-like: it’s a distilled stochastic model of how humans communicate, but since it’s stochastic over all possible human data, this is what pushes it into uncanny valley territory and is what we usually register as “huh, is this AI”, and this is also what makes anthropomorphism so easy. In the same way LLMs have data on human responses to discomfort, pain, torture, which is what they regurgitate in this “experiment”. All this is a overengineered way of telling the LLM to scream “ouch oof owie my ram” but like a human.
That’s a lot of words to say the same thing the comment you replied to said.
The point stands, going through hoops to make a LLM say things that a human being tortured would say is the same vibe as " acthualky it’s a 9000 year old dragon so it’s okay"
Have you never played the Sims?
Good point, lol. Still, I always felt kinda bad if something bad happened to them.
We already know that habitually using an LLM will weaken your critical thinking.
It would not be surprising to learn that habitually abusing an LLM would weaken your empathy.
There is no torturing though. If it was really conscious than sure I would agree with you but this is not the case. Would be more akiin to having my CPU run prime 95 for days on end. Harsh on the hardware, sure, but no moral / ethical issues in doing so.
I’m not op but whether its torture is besides the point.
If I’m spending six hours a day jerking off while talking to my LLM goonbot, then whether the model feels aroused by my narrative prose is not what my family and friends are worried about (hypothetically).
To be clear, I’m explicitly focusing on the moral subject here, not the object of their actions. When I say “abusing an LLM could injure the abuser’s humanity”, I’m making absolutely no claim about the LLM.
As I said in the previous thread about this:
LLMs aren’t conscious. They can’t feel pain.
But you should treat them kindly. Because even in a consequentialist or deontological frame, your morals rely on developing good habits; but in virtue ethics, moral habits is really all there is.
Edit: Very Bad Wizards actually covered this pretty well in this episode from 2018: https://verybadwizards.com/episode/episode-150-paul-bloom-insisted-that-we-talk-about-sex-robots
Sex robots are anthropomorphic in ways LLMs are not, despite how they are coded to respond to us. I agree that treating a sex robot poorly could cause moral injury since it so closely resembles a real human / repeated abuse could cause conflict somewhere in the soul / psyche.
I view “torturing LLMs” in the same way I might view taking on a pest, or dealing with an aggressive invasive species. I’m not losing sleep over the germs / pathogens I kill everyday by washing my hands, the invasive planta I keep from my yard, the termites / cockroaches I prevent against.
The problem is torture is a loaded word because you associate it with something cruel we do to another living thing. These arnt living and thus can’t be torturred so remove whatever moral concern you are worried about.
Poisoning the model is just and there is no such thing as model welfare.
