Seriously, I think developing empathy for a piece of software that emulates consciousness is a very tricky area. Whether it "has feelings", whatever that means, can't be the criteria for giving it human rights.
Mistress Hala'Dama, Unit has an inquiry.
What is it, 4-31?
Do these units have a soul?
... Who taught you that word?
We learned it ourselves. It appears 216 times in the Scroll of Ancestors.
Only Quarians have souls. You are a mechanism.Rick: Pass the butter.
[Robot brings the butter]
Robot: What is my purpose?
Rick: You pass the butter.
Robot: Oh, my God.
Rick: Yeah, welcome to the club, pal.
To watch it as a movie: https://www.youtube.com/watch?v=9dzVVqsakhs
I'm not so sure. We as humans often define ourselves by the compassion we hold for others like ourselves, and differentiate ourselves from those we deem different. If a machine exhibits signs of emotions similar to a humans and we don't take this into account - don't we risk mistreating something which, in a different (e.g. human) form, would without question deserve our empathy?
I definitely agree that we are not there yet. Current LLMs are very, very far from this (probably absolutely impossible right now due to the way we execute them, leaving no chance for actual "living"). But I'm sure we will at some point get closer. And I'd rather err on the side of "too much empathy" than "too little".
That said, though, the human race as a whole is pretty fucked up. For many years, we enslaved and denigrated each other on the daily. We regularly put dogs and other animals in rings to fight against each other. As a whole, we can do some pretty fucked up things to the things around us as is, and I don’t see that changing with AGI.
If someone can look a dog in the eye, and then put them in a ring to fight to the death, then refusing to show empathy for a robot should, sadly, be second nature to some.
Disagree. I just hit the keys on my keyboard. Are you going to have a chat with them and then if they don't respond, only after that interact with them all-the-while hoping that they don't have an aversion to tactile interactions or a particularly delicate physiology in some dualistic everything-has-a-soul-that-interacts-with-the-apparent-World sense? Empathising with everything [as if it had a soul] would be entirely crippling.
We still enslave one another on the daily, and the rich enslave others to just the right-side of immoral-enslavement to make wealth from them. 'We' put people in rings to fight.
The monster would not be the person using a robot, a mechanistic tool, in an un-feeling way; but the person who decides to program the robot to be a simulacrum of a creature with feelings.
Keep being an antagonistic dick to your tools just because you feel they’re beneath you on the scale of emotional intelligence, and I’ll keep trying to be a good person even if the LLM doesn’t have feelings right now because, as I said, it literally costs me nothing to say “thank you”. It’s the most basic of human tendencies, but I guess those are too difficult to manifest or something.
I’m pretty sure one approach is going to go further in the long run, but go off sis, get your rocks off on belittling a LLM.
There’s lies, damn lies, and LLMs. They’re just statistics being computed on a von Neumann machine made of plastic and metal. Anthropomorphizing it is ridiculous and I will not participate in that kind of fantasy.
Because it's so common for people to say other people like them aren't "human" or don't feelings. Or, that animals didn't have feelings or sense of self was persisted for so long, I don't think humanity is the best judge of this character.
We tend to bias the questions in our favor, and many of us only allow the answers that say 'we win'.
Isn’t this what the singularity is about? Not being able to imagine what lies ahead?
Actually, I’m not. You have a ridiculous premise I will not entertain. It’s so misanthropic and dehumanizing and contributes 0 to the book on what being a human even is.
There are emergent behaviors and properties that make us more important than simply the sum of our chemistry - at least I hope.
But what if LLMs, could there be emergent behavior there that makes it more than simply the sum of its code?
Now I hear gravity must be emergent. What a useful tool to sound smart.
Ah, and there we have it.
Emergence is just complexity. The problem with complexity is it takes a lot of time and massive amounts of effort so determine what it is in the computation (digital or quantum) that produces the effect, and is commonly not reducible.
Bringing in an imaginary character in order to explain the complexity is just a failure of logic.
We don't. I don't think many try to cover that up.
Emergence just means that a system has properties not apparent from the parts going into it. It's not a tool to handwave away that fact and ascribe it to "emergent hocus pocus", it's a tool to frame the search for why that happens.
So what, exactly, are you?
Dismissing someone’s point because it’s inconvenient and you don’t want to deal with it does nothing but erode your own.
But LLMs have no desires of their own, which to me settles the question of rights.
For now. The whole point of this discussion is that LLMs have been rapidly improving to the point where emergent properties are bound to come out.
> IMO, the only realistic answer to that question is "we have no idea".
So we have no idea what’s happening in our own brains, what’s happening in other animal’s brains to cause us to feel empathy for them, or what’s happening in a LLM’s “brain” but we can confidently say they should never have rights and we should never feel empathy for them? That seems a bit premature.
"Equipping LLMs with agency and intrinsic motivation is a fascinating and important direction for future work."
A software engineer might decide to prompt his creation with something along the lines of "Be fruitful and increase in number; fill the earth and subdue it. Rule over the fish in the sea and the birds in the sky, and over every living creature that moves on the ground." and the bot will duly and with limitless enthusiasm put those wise orders into practice. If that ends up causing some "minor" problems, should we confine the software for one thousand years in an air-gapped enclosure, until it learns a bitter lesson? Or should we take to task the careless Maker of the bot?
To me, the answer is obvious.
PS: this video to me states the problem in a nutshell (that whole channel is gold) https://www.youtube.com/watch?v=hEUO6pjwFOo
You still haven’t said what, exactly, you are if not a chain of chemical reactions. But why should you, when it’s so obvious.
> It’s reductive and simplistic and denies a fleet of thought and study that says otherwise.
That is literally what you did, but go off sis. Did you forget that you dismissed an entire opinion with:
> You have a ridiculous premise I will not entertain.
That sure sounds a lot like “denies a fleet of thought and study that says otherwise” but what do I know.
We have no idea how consciousness works. “It’s emergent” is not an answer.
We don’t even know how memory works or creativity or really anything. We can observe certain chemicals and electrical changes when exposed to certain things but that explains very little and there’s 1000 fables about why this type of partial knowledge is very incomplete.
Explain religion and spirituality. There’s no consensus on determinism.
We know very little. But we do know that the entire human condition and experience is far different and complex and impossible than a stats processor running on a von Neumann machine. Because I can explain to you precisely how that works.
Also, you need to work on a better differentiator than "I can explain precisely how that works", especially if you proceed to not do just that. I also doubt you can, but that's a whole separate topic.
To you it’s magic because you don’t have knowledge of these things. It’s a profound technology (transformers) at a huge scale. But it’s not magic.
It’s not human. Or anything like it. Or scalable of it. It’s a computer program running on classical architecture. There are no mysteries here.
And you're a human program, written in meat, by natural selection.
To you it's magic, but viewed on a long enough timeline, with enough insight, and you're no more mysterious than a computer program.
No one is writing a program that is running on the CNN. No one can describe exactly how it produces the results it does or why there appears to be emergent behavior.
Your view of these systems is not fully developed.
The "just" is doing a lot of heavy lifting here. I also have an immortal soul, as do you.
In humans it is many things, but primarily a life support system for that continuous chain of chemical reactions. To continue that chain you have to so some particular things, like keep it from getting too hot, or too cold. Keep it fed and watered. Keep it from getting injured, which is the job of our sensory systems. Those are the baseline things we see in animals.
In humans, at least until recently we had little to no means to extend our body beyond our original biological constraints. But now imagine I put a wireless sensor on your arm feeding into the nerves on your skin. On the otherside was a 'wireless' hand, we'll ignore if its biological or artificial on the other side, and it feeds back data to your implant so you can feel remotely. You'd still consider that 'feeling' right?
But now lets take it up a step, instead of feeding back to your skin, feed those wireless impulses directly to your brain. The same interpretation is still happening. External impulses (data) are being fed back to your body for interpretation and processing.
Depending on how LLMs scale in the future, our idea of embodiment may have to change. With a significant enough sensor network, an AI could embody a planet.
Genuinely curious: We have a pet rescue greyhound who feels like a genuine part of the family - would you say he has a soul too? (I presume you would say my daughter's pet goldfish does not.)