Hacker News Reader

Comment by card_zero | Hacker News Reader

Do LLMs grieve?

You can give them emotional signals, but I feel like it's cruel to do so and would recommend against it.

One of the important overlooked aspects of this is how maltreatment of something that appears to be human-like in many ways by an actual human affects the person doing it.

I know this did not end up being applicable in the video game world (which I'm continually glad for). With LLMs, it feels more likely it would be.

We’ve had 30 years of people killing digital people, often controlled by actual people without any obvious impact on violent tendencies.

I’m concerned about campaigns by misguided people to “protect” the bots. Seeing the response to the Unitree bots on AGT and their behavior’s effect on my own emotional response, I wonder how far off we are from calls for robot rights. We are more than tool-using apes, but for those that think that’s all we are, what’s the difference between that and tool using bots that emote, mimic distress, and otherwise act autonomously.

This is one of those things that really does need checked up on occasionally.

For example, do video games not cause violence because of something like being on the wrong side of the uncanny valley. We know it's fake therefore its fake.

Do does this break down when it gets "too real".

For example lets imagine a future VR video game that the inside game doesn't feel much different from the world we live in? Lets also imagine an interesting unique constraint in this video game. If you die in it you cannot play it any more. You play it for years safely and make romantic connections in it. One day someone kills your partner violently in front of you. You'd be pretty willing to take up violence against said person, right? How much of that violence would carry out of the game?

I don't want to make proclamations like the video game violence people did that were not correct, but I also don't want to assume as we make our simulacra more like us that it won't affect us somehow. Situations like this are rarely binary, and it's best to keep an eye on it.

Lastly, because LLMs can and do effect the real world with their actions, it's likely a good idea to error to the side of caution and at least be polite.

Insofar as grieving is an inherently, fundamentally a very particular chemical process, the answer is a resounding “no”.

If you simulate that chemical process in a computer, if necessary a quantum computer?

You don't even need quantum imo.

On a personal meatbag level emotions are magical and something ineffable.

From a scientific stance, chemical make brain go happy or sad and there's nothing special or magical about that. Every emotion we feel is because it was evolutionarily valuable for us to do so.

This is too mechanistic. You also choose which emotions to pay attention to.

This is too simple in itself.

When your afraid you can choose to ignore the fear, but you are still afraid and it has physical effects that can be measured. There is at least a trinary state of possible outcomes and as the situation space becomes more complex and solutions drift apart on all 3.

Then, logically speaking, you get a simulation of griefing.

And in what lies the difference?

That's a very closed definition of grieving. So by your definition a future perfectly simulated human brain cannot grieve because no chemicals are involved?

Lmao.

Whether anything a perfectly simulated human brain (or other organism see the fruit fly simulated brain) does is actually real - in the same sense it’s real for you and me - is still up for debate i think.

Assuming it’s possible to perfectly simulate a human brain

A “perfect simulation” is no longer a simulation.

The definition of the word “simulation” is:

> A simulation is the artificial imitation of a real-world process, system, or event using a model

As we know, “all models are wrong; some models are useful”.

A “perfect simulation” is, then, an oxymoron.

Thanks for the semantics, a perfect imitation would be a better description

do humans? or is it only because we're the ones doing the research and so we say we do?

i think most people are automatons and can be programmed.

it is the most moral thing to program people to operate in ways that causes least harm to everyone else around them that has to live with them.

people like certain foods, outrage, and gambling/fomo.

> do humans? or is it only because we're the ones doing the research and so we say we do?

...literally on this article.

Do androids dream of electric sheep?

Nope.

They can pretend to, of course, and use the language of grief, but if you can't feel the cold bottomless pit that has bored through your gut then you aren't really grieving.

Do whales grieve?

> Do whales grieve?

Just as much as any human that isn't me, as far as I can tell from all available evidence.

So, yes, I absolutely believe they grieve, just like I believe humans who aren't me grieve.

I hope that’s just because you haven’t had a reason to grieve… Either way, hope you are well!

> I hope that’s just because you haven’t had a reason to grieve… Either way, hope you are well!

Thanks, stranger.

I'm old enough that I have experienced significant grief a few times in my life.

Given our mortality, we should welcome some degree of situational grief, as grief is the absence of a previous love that we were lucky enough to experience in the first place.

Of course, there are also profoundly unfair versions of grief (like the loss of a child) that I wish nobody (or no whale, for that matter) would ever have to experience.

"That isn't me" in the above comment means "I know I can grieve, but I can only tell indirectly other people can do so as well."

It doesn't mean "I've never felt grief".

They exhibit behaviours that seem to defy alternative explanations, though I doubt anyone could say for certain whether they do or don’t

You could have an alternative explanation: large animals invest lots of food energy in their offspring, and so the slimmest chance that a calf might not be dead is worth lots of effort. Say one in a thousand actually revive.

Here's an article about orcas: https://www.bbc.co.uk/news/science-environment-64559047

It seems that an orca invests so much in a son that it's like an all-or-nothing one way ticket to pass her genes on. (She might get second try later, but the probability is much lower.)

You might also argue that although there may be a non-emotional argument for why grieving could be beneficial, it’s possible or even likely that evolution encoded this instinct in something like emotions. Just how evolution encoded the tendency to care for one’s family and community in emotions for humans.

Yes! Which leads into strange territory, because we can say that attachment is not entirely rational, but is influenced by genes pointing us toward more successful reproduction, without concern for creating knowledge. But then this environment of instincts is the environment we've built up all our cultural store of meaning in, it's the human condition, so the argument can be turned around to say that meaning isn't entirely rational. So even gene-driven attachment isn't meaningless, but at the same time, I don't want to be bullied and herded around emotionally by little nucleic acid spirals just so they can replicate themselves.

> They exhibit behaviours that seem to defy alternative explanations

No, they really don't. The article says

> indicated an inherent, strong maternal attachment to the calf

Yeah, that's not grieving.

What if I told you that marine biology is a large and rigorous field that is not fully encapsulated by TFA?

Indeed. The same goes for widows at funerals. I mean, has anyone ever dissected the brain of a grieving human and said “aha hah! Here it is!! The grief!”

No, of course I haven't been doing that, and it's going to be a while before I would have collected a representative sample anyway, if I had been.

How do you know that they don’t? And conversely, how do you know any other human than yourself do?

Do books grieve? I ascribe the same kind of interaction to llm's. They work on a different substrate than us and their perception of reality, if there is any, will be wholly different than our perception because of the constraints of the substrate itself.

Not to mention the lack of continuity of existence kind of precludes any deeper and more meaningful interactions with the self.

I am not saying that they are unable to feel, but at the current structure and ephemeral existance they are unable to do so, they lack the minimum prerequisites.

>Do books grieve?

Is there any ordered state changes occurring in books?

>they lack the minimum prerequisites.

And these minimum prerequisites are? There are a number of scientists that would love your empirical evidence.

>lack of continuity of existence

Eh, the lengths of timelines that LLMs can execute for, and this is in singular execution and agentic evolution is ever increasing. There isn't any time based requirements either. For example I can give you all kinds of drugs that turn off and on your conscious awareness and this doesn't stop self interaction when you're awake.

Reply on news.ycombinator.com