I’m still waiting for a Goatse powered LLM.
I think theres a big hole in that plan.
It would be quite a stretch.
Yeah that’s a real prolapse of judgment …
Are we married to this idea?
LLMs are not sentient, but this gimmick doesn’t prove that.
There’s nothing to indicate the human brain can’t be implemented in a Turing machine too.
There may not be proof that LLMs aren’t sentient, but there is some evidence and reason to believe that consciousness is an emergent property of quantum effects, whereas turing machines are relegated to classical mechanics. The same evidence would indicate it is unlikely that a turing machine could replicate the human brain since it uses fundamentally different layers of reality to calculate/think.
it baffles me how many people in here apparently don’t have the reading comprehension to understand your second sentence. I swear Lemmy is even dumber than reddit.
There’s nothing to indicate the human brain can’t be implemented in a Turing machine too.
That is not true, this is a research branch from computer science and math, it is called Computability theory, it deals with the limits of expressiveness for different types of theoretical machines and expressions, and the most expressive of all is the Turing machine, and a Turing machine cannot do some stuff, the classic example is the Halting problem, a computer cannot definitely say if an algorithm ever stops (mathematica proven that it cannot do it), but a human can do so quite easily.
One may think that maybe a Turing Machine cannot do something but can simulate another machine that does, but that is also proven to be impossible, it cannot simulate something more expressive than itself.
I know computability theory, and I am very familiar with the halting problem. A human cannot solve it either. We made literal mathematical proof of it, and that proof is the halting problem.
The entire point of the halting problem is that if you assume that there is a black box that can answer whether a program halts, you then prove it can’t be the case by a proof of contradiction. You can replace the black box with a human brain and it works just as well, that’s the entire point of a black box.
You’re misunderstanding the halting problem.
You are refuting something that wasn’t said. of course Turing machines cannot compute everything.
I don’t think that is the case.
There’s nothing that indicates the human brain has super-turing computation
Except that we’ve never done it, and we struggle to even define what consciousness is. But, other than that, sure.
There’s fundamentally not much difference between our brain and a fly’s, at the cellular level. We have fully simulated a fly’s brain already. When given a virtual body, it promptly started acting like a fly.
I don’t think LLMs are conscious or sentient. However, consciousness is likely just an internal illusion. There’s no obvious reason we can’t scale up from a fly to a human brain, other than difficulty. At that point you have a fully virtual brain that believes itself to be conscious, and can demonstrate sentience.
I’ll believe it when it happens. There’s a huge different between simulating 139,000 neurons and 50 billion neurons.
Besides, the authors looked at taste and touch sensory. That’s not exactly predicting the entire behavior of a fly, and the authors (of which there were many), admitted they don’t know if it can actually predict neural activity. Source
I’m not expecting anything any time soon either. Though I can see someone like musk pumping far too much money into it at some point.
My point was however that the difference is just one of scale. We don’t need to predict the firings, just run it and compare it to nature. From what I’ve read, it behaves like a fly, including walking and grooming itself. This means there is no magic mystical difference between a real fly’s brain and a virtualized one.
Projecting further, there is no difference, other than scale, between our brain and the fly. Implying there is nothing mystical about consciousness.
If a human brain can be conscious, then a virtualized human brain can be conscious. If a virtualized brain can be conscious, then so can the computer it runs on.
The question then becomes do we WANT consciousness in an AI, what would it look like, and how can we detect/measure it?
The difference of scale is in the grey area of the difference of scale of quantum and classical mechanics, though. Conciousness very much could be something that depends on the emergent properties of quantum mechanics and doesn’t reach classical mechanics.
This means there is no magic mystical difference between a real fly’s brain and a virtualized one.
That’s not true. You might simulate fly intelligence, but you aren’t simulating a fly. Flys are an entity driven by their phenomenological experience of the world (as all us nervous systems are). It would be a rather strange thing to say that a fly is only the pattern of behavior you recognize as a fly’s behavior.
Note that flies also have this capacity to self-evolve over generations. There is no single fly that you can point to and say, “that’s the right one. Copy It.” So, your “fly simulations” are always at best a behavioral approximation. Given enough time, say a decade or century, this ought be obvious by the fact that your simulation no longer accurately resembles the most modern flies.
Projecting further, there is no difference, other than scale, between our brain and the fly.
That’s just not true. Consider “What The Frog Eye Tells The Frog Brain.” Put briefly, the eye encodes and transmits semantic information — as opposed to the more common belief that it transmits raw visual information. That said, there a trillions of differences between how that might work in a human versus in a frog, let alone a fly.
I think you’ve discovered “neurons work similarly across species,” which is like saying “thing does same thing when used in other location.” This doesn’t tell you how neurons work to drive that behavior, it doesn’t tell you what it’s like to experience your neurons firing that way, it doesn’t tell you why the neurons were developed that way topologically over time… At best, it helps you develop a timebound understanding of fly automata to neural architecture. That’s without any understanding of phenomenology to neural architecture, and likely without being able to decompose the neural system into any semblance of semantic information processing.
A human is much more complex than a fly. I can’t believe for a second that this approach would work scale to a human’s brain. If you captured and simulated every neuron in the human brain, you’d be left with a feedforward process simulating a particular neural state of a particular person. No feeling, no reflection, no introspection, no semantic meaning. Just a neat toy.
Buddy, I barely think humans are sentient
An illusion is a phenomenological experience, for which you must be conscious. Consciousness can not be an illusion. You must be conscious to experience an illusion at all.
It’s a case of running out of terms. I was using it here as apparent, but not a “real” thing.
It’s a bit like a lot of visual illusions, we can often all see them consistently, but they don’t exist in the image itself.
In this case, consciousness is likely related to keeping our own mind functioning coherently. Providing a common virtual ground for the various parts of our brain to interact. There is no seat of consciousness, it’s akin to the operating system on a computer. Not required, but makes a lot of tasks massively easier.
If that’s the case, by calling in an illusion, aren’t you just stating that it’s a misunderstood process?
Consciousness absolutely is something. That might be a procedural “thing,” and I think you’re saying that the notion of an ontological consciousness is a misconception. An Enlightenment Era misconception, if I’m allowed to add that in there…
To call it an illusion comes across as though you’re asserting what it is. I think you’re really just pointing out that it isnt what a lot of people consider it to be (often implicitly).
Is that fair?
A simulation that isn’t aware that it is a simulation is not self-aware, which is a qualitative difference from what is being simulated. Consciousness isn’t like some rom you can emulate.
You say that with some confidence. What makes being aware of being a simulation a binary criteria for self awareness? You may just as well be in a simulation and probably see yourself as fairly self aware.
Sure, and how do you know your thoughts are your own? Like if you have to defend your position by questioning the nature of reality then you got bigger problems.
It is. There is limitations to a Turing machine that a human brain does not have. I expand on that on this other comment here.
Except … consciousness!
Which you can precisely define, right?
What a moronic take, we cannot prove consciousness (except our own of course) so it’d be you who must prove it can be implemented in a turing machine. Not me.
Statement: “There is no indication that the human brain cannot be modeled as a turing machine”
To disprove this, evidence to the contrary is required. It’s not at all the same as saying “The human brain can be modeled as a turing machine”. In that case they would need to prove that.
We simply do not know. Humans cling to their idea of somehow being “special” very hard with thought experiments like chinese room, and at all points neglect that there is no evidence that a human brain is actually different.
Generally I’d argue that the continuous nature of animal brains makes them quite fundamentally different from the very much discrete states of anything we program, but that still doesn’t mean it’d be impossible to simulate.
Except we have quala, consciousness. That’s a well recognised fact. How do you model something we cannot explain?
So yeah there are “indications” we cannot do it. Maybe we can and emerging properties is the way to go, but it might also not be that.
Do you believe we can?
Edit: the chinese room is a farce IMO, and won’t push anything either way.
Many abilities and properties we thought to be exclusive to human consciousness were later proven in animals. Such as the capacity for empathy, thinking ahead, and recognizing yourself in a mirror.
Does that mean animals are partially conscious? Or can we accept that the definition is fuzzy at best, making this a moving target?
That makes it probable I guess, but it absolutely does not prove it. We cannot prove humans are conscious (except ourselves).
By the same logic, you could prove that we are not sentient.
Engineers need to stop trying to do the job of philosophers. LLMs are not sentient, but that tells us nothing about what gives rise to sentience or whether silicon is a substrate that could host a sentient consciousness.
I’m still not convinced most people I speak to at work are sentient.
I’m pretty sure many of them, when they get home and close the door behind them, they just stare at the wall until it’s time to go out again.
big issue is that “Scentience” is a vibes based definition. I don’t think we have a working definition that we can use.
in the past it was argued (engire global economy based on) that black people weren’t scientient.
The annoying part is that now, “what is scientience” isn’t an abstract esoteric philosophical question. but a material question we are ignoring with serious real life implications.
They are probably sentient. Now sapience… is another issue altogether.
I majored in Philosophy. Though I wouldn’t want to be an AI ethics consultant for one of these big companies because I imagine that whatever I write down would have to preserve profits (first and foremost) and then gaslight everyone into thinking it’s all perfectly ethical and okay and no one (including mother earth) is being harmed in the process.
You don’t need an ethicist if you don’t have ethics! /s
I just wish we could put this dumb sentient AI stuff to bed for now. Congrats, you guys built the most powerful (and wasteful) auto-complete. It’s not sentient and your marketing doesn’t have to be hyperbolic lies.
honest question to a philosophy major, why is it so clear cut to you that LLMs are not sentient? what is missing in current systems?
Protip, never ask a Phil major questions, unless you want them to pose more questions. I think the very question itself assumes we can even properly define sentience. If we can, why does it have value as a result? Your home computer may not be sentient, but it’s pretty smart, valuable, and could potentially have an equally large impact on the world (not sure about intentionality though). It’s not that I’m certain no LLM is sentient (though I’m pretty sure) as much as I see greedy people steering the hype to their benefit in order to exploit others. The whole thing is manufactured to cheapen humanity in comparison to tech. I am very opinionated so you may see it differently.
Can you build a human from goats in AOE2?
For this proof to work, you’d have to prove that you can’t. And we can simulate whole worm brains and fly brains, and they behave like real in simulated worlds, so you’d have to prove that there’s something different about humans. This paper doesn’t offer anything like that.
An average human? No. A lemmy.world user? Yes.
Yeah, fuck those idiots for joining when it was the only instance accepting new members
Haha, I’m still trying to decide if the upvotes and downvotes here are going after the sarcasm or the dumb literal take. Snort
I have no knowledge of this drama or hard feelings for your instance in either direction.
But was that ever a thing? I joined a month after you and there where options plenty.
It was absolutely a thing right when the 3P app reddit fiasco happened. Everybody was scrambling, and I’m not surprised that there were tons more options even just a month later, but at the time everyone else was completely overloaded. I tried signing up for as many instances as I could find, and lemmy.world was the only one that wasn’t shitting the bed with new registrations.
I just found out I’m scum on lemmy woot! Who knew people didn’t like certain instances, are some better than others? Like do you get better info faster? Or is this pure sarcasm?
Are we discriminating against instances now? Is this a thing?
Sweetie, that was a thing before your instance existed.
For the record, at least one Nobel Prize winner disagrees about LLM sentience
A lot of nobel laureates believe/believed a lot of unhinged nonsense. Some of my favorites:
-
Linus Pauling thought overdosing on vitamin C could cure all kinds of medical problems
-
James Watson and William Shockley were eugenicists
-
Irving Langmuir believed large-scale weather manipulation was possible with the technology available in the 1970s
-
Luc Montagnier was really into homeopathy and spread COVID vaccine/5G conspiracy theories
-
Kary Mullis denied that HIV causes AIDS
Just winning a Nobel doesn’t automatically make someone right about everything, even in their own field.
-
Exactly!
The researcher didn’t make an LLM, he made a NAND gate, and in theory that could be expanded into an LLM
All computers are just a minimum of three layers of NAND gates
he made more than a NAND gate. he created a 1-bit perceptron and a perceptron-training circuit after figuring out how to make a NAND gate
That is still a long way from an LLM
I see the point that the LLM is closer to a 1bit perceptron then it is to the brain?
Lost me at Microslop
Does it matter who his employer is?
The fact that he did this and wrote a paper on it while MS continues to pour billions into LLMs actually commands respect.
Undermines credibility for me.
“Man built brain out of simple particles, proving brains are not sentient.”
Speak for yourself. I’m a Boltzmann brain.
HOLY SHIT, AGE OF EMPIRES II IS SENTIENT
Who knew age of empires was gonna one up the president? In this the year of our Lord 2026
Lord Henry Percy: Thou art human, with soul and wit! I am naught but clockwork!
Lord Henry Percy: No wonder thou wert victorious! I shall abdicate.
Lord Henry Percy has resigned.
The moderate difficulty computer I just beat disagrees.
It is incredible how we forgotten the whole point of the Turning test the moment it became relevant rather than some abstract though experiment.
People it this thread does not know what the Turing test actually tests. It is Ok, I guess the blame fails on pop culture.
The Turing test is not a test of sentience, it is a test if a machine can fool a human being into thinking it is sentient. And machines have passed this test for decades already. Who would have thought that our predisposition to anthropomorphize everything would make us so susceptible to thinking stuff is sentient 🤷
it isn’t just a test of competency. but the philosophical questions that are brought up afterwards.
It’s like reducing Schrödinger’s cat into an animal abuse experiment and not what it means to have uncolapsed probably waves.
The Turing test doesn’t account for state. LLMs, while they could pass that test, are idle when unprompted. They dont have a means of responding to any stimulus but those provided. If they were provided even a fraction of the stimuli provided to a real mind, they would rapidly consume all available system resources trying to respond, regardless of how many we could reasonably provide.
Also, they are fixed. LLMs do not change once put together, and only seem to based on a rolling context window they store based on their previous interactions with the subject. They cannot internalize any of that interaction to change their underlying model or its weights.
Because of these things, I believe it illustrates how the Turing test, while an important thought experiment, is incomplete regarding defining a thinking machine and the ethics surrounding it. If the machine is off if I’m not directing it and can’t functionally remember or experience anything, it can’t experience suffering or oppression or any of the things associated with its agency, freedom, or any of the philosophical underpinnings of what constitutes another entity.
The turing test is not relevant anymore. Any LLM would pass the turing test without much problems.
Go read what Turing proposed, it is a lot more complex then how it is often parroted. There are many more tests that the latest models would not pass, like explaining poetry (that it was not trained on)
it is more relevant than ever.
it wasn’t a test for sentience, but a thought experiment on what sentience is.
and now that it has been passed, we need to seriously consider what is or isn’t scientient.
and so far, haven’t heard anyone provide a good definition or test. or even guidelines. and more importantly, laws.
if a lab was to create a scientient LLM, are they allowed to turned it off? shouldn it be illegal to try?
How tf can you define sentience?
Hate to be that guy, but it’s “sentient”. It has nothing to do with science or scent.
hate to be that guy, but since we created tools that can pass the Turin test. scientience is part of science.
btw, your allowed to have unsolvable philosophical questions at the core of your field.
Biology has no definition of what life is. we biologist just gave up trying to define it. the definition they teach is highschool is BS.
… I was just pointing out that you keep misspelling it.
Sorry, spent the first year of my PhD learning to spell senescence.
My condolenscenes.
Any LLM would pass the turing test without much problems.
Lol no. There are soooo many completely terrible bots. The vast majority.
that’s the whole fucking point of the Turing test.
that sentence isn’t a strict true or false, and if you can’t tell the difference, then the real question is wether there is a difference.
maybe the only ethical solution is to assume it is and consider using it a form of extremely unethical slavery.
under that lens modern AI is some rick and Morty bullshit. Like imagine creating a life for a menial task then executing it as soon as it’s done. Then everyone discussing it’s competency rather than how fucked up it is.
I’m not grasping what you want to say.
The turing test is an outdated test, it has been an important milestone but nobody working on computer sentience either philosophically or scientifically considers it a valid test anymore.
According to the turing test, LLMs are sentient beings. The widespread opinion is that they are not.
sentient beings
I don’t think the test was ever meant to show something this strong.
According to the test, it might be. The biggest reason why it isn’t is according to the capitalist financial interest.
I’m not saying it is. or isn’t.
but we cannot just with this toys that we don’t understand.
To be fair, humans anthropomorphising everything, of course they do that with software that can fucking talk to them.
John Searle is the GOAT?
