this post was submitted on 23 Aug 2026
76 points (82.2% liked)
Technology
87550 readers
3298 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 3 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I mean... No? Absolutely not.
I dont even know where to begin. Maybe with their state being fixed in time; LLMs do not change. An sci fi analogy might be the "no timers" in orion's arm, who are but a single thought frozen in an infinite loop in time:
https://www.orionsarm.com/eg-article/47f4311eaef31
Except they arent even that, because its just a next word completion model, not something that thinks. It is not self aware.
https://arxiv.org/pdf/2503.09211
https://openreview.net/forum?id=klU4737opt
https://arxiv.org/abs/2504.09762
This becomes (to me) very obvious if you ever use an LLM in raw completion mode. It very smart, but at the end of the day its no different than a weather prediction model spitting out probabilities for a storm system.
Chat finetuning is meant to get humans to anthropomorphize them by training on human preferences, and the interface further reinforces this. Its all a trick, albeit a very elaborote one.
Will future architectures be closer to "thinking?"
Maybe.
But we are a long way away.
Not with agents and chain of thought. Agents can run for hours continuously. So sure, an AI agent is not the same as a sentient human on the time scale of a year, but on the time scale of a few hours, perhaps the AI is "sentient".
The first paper you linked is focused on the mechanism and not the output. It creates a definition of "thought" and then talks about how the AI doesn't "think in the feature space". It only addresses chain of thought at the end, and says that the issue is that it constrains AI to think in natural language only (instead of, say, pictures).
But again, why would thinking in pictures define sentience? The paper gives a rock-paper-scissors example and says that the AI thinks about it in a different way than a human. So what? If a human plays rock paper scissors against the AI and the AI's output is indistinguishable from a human's, 99.9% of the time, why is that not sentient?
This is like saying python programmers are not real programmers. If a python programmer can implement the same program in python, who cares what language they use.
I'd love for a better definition of sentience than "it works differently than humans".
Them
You
Those are not the same concepts. The model of LLM remain unchanged. Agents running for hours just modify the prompt they input into an LLM model.
I do not care about the rest of discussion, just swoop in to clarify that distinction.
Fair enough. When I say that AI might be sentient, I include systems like "LLM + agent harness", treating the whole system as AI.
That's too much for their tiny brains to handle.
The fact they are trained to complete sentences and are thus word predicting models only tells us vaguely how they work. It does not tell us anything about their sentience.
The sentience, if there is, comes from the emergent reasoning that develops in the massive neural network to offer actual good predictions: the more you think, the better the predictions.
For example, models trained to produce textual board games predictions can be sounded to show they developed subnetworks corresponding to an inner representation of the states of the game. Models trained to produce images have emergent subnetworks on depth-maps, contours and lighting.
Are cells sentient? Bacteria, fungi, etc. Do you think they have any degree of interiority, like how simple animals do?
Exactly. It's not so cut and dry. Are they? You can never know until you experience their existence.
Lol no. Actually cells are demonstrably NOT sentient. They have no capacity for choice. No capacity for emotion or experience. They are literally molecular machinery. And we(scientists) understand that machinery to a shocking degree.
If you think cells have any degree of sentience, then you must make a similar argument for advanced clockwork machines. Not robots, pure wind-up clockwork automata.
Ah, so you've experienced each of their existences? Fascinating.
Please explain to me how a wristwatch is sentient or conscious.
I can't, since I've never been a wristwatch. Maybe they're not, or maybe they are through a mechanism for which science has no current explanation. We can't know, only surmise.
Are we not just prediction systems with more neurons?
Now im not saying LLMs are human at all because to be human is to have gone through our exact evolutionary path, but we really are in a lot of ways much more similar than you think. At least the speaking part of our brain is.
The argument that LLMs are fixed in time does not contradict consciousness. Say a human was stuck in a time loop and the human is asked the same question every single loop, at the exact same time, in the exact same way, with every single atom being in the exact place. Would their answer change? I would argue no, the same exact neurons would be stimulated every time that would cause the same exact answer every time.
We don't even have a solid definition of what is concious. Are insects conscious? Like the fruit fly, we have modeled all of its neural paths and it simplifies down to essentially being a robot that responds to stimuli.
What does make humans special is our ability to learn. But what when an LLM trains its self? Is that not learning? I think the question of consciousness is too complex to simply say one way or the other and it is foolish to simply it this way.
Thats what I'm saying. They cannot do this, not even close. Even a fruit fly is somewhat adaptive and can respond to novel stimuli, but an LLM effectively cant.
There are tiny experimental models closer to a fruit fly than large LLMs now.
...What about future world models with some adaptive learning loop, sampling stripped out, and so on?
Yeah! That would be fascinating. Then sentience starts to become more of a question.
But thats not what LLMs are.
People like Altman and Modi have grossly overexaggerated what current LLM architectures are capable of, they trained the LLMs to reinforce the illusion as much as they can, and they arent to inclined to change it. Hence all the good researchers have fled to research world models, and are saying transformers LLMs are not a viable path forward.
Even in your highly optimistic scenario, the AI wouldn’t even come close to being comparable to actual sentience.
Well, theres a lot of experimentation with actual neural nets mimicing biology (or literally using biological neurons), and I think that could start to raise the spectre.
World model research could be a significant component of that.
In principle, OP is right; humans are just machines. All I constest is that transformers LLMs are even in the same ballpark, or possibly a path there... they are not.
Well it sounds like we agree then.
I would think you would need actual biological structures to achieve sentience. That’s a lot different than what these people are calling AI tho.
Edit: Also, if it is built out of natural biological pieces, at what point does it cease to be “artificial”?
Well biological structures can be replicated mechanically. So at risk of being pedantic, that's not strictly true.
I view "artificial" as just a term for conscious, purpose construction, whereas "natural" does not involve any conscious intelligence, as in "naturally occurring in nature." But even that is a spectrum, I guess, with genetic manipulation, BCIs, all sorts of intermediate possibilities.
I'd call a constructed, purely biological brain "artificial" because, even if it was grown naturally, it was built with conscious intent.
Correct, nothing is special about biological consciousness. If these exact functions are replicated mechanically/artificially, the outcomes would be the same.
I can’t believe how many pro-AI arguments essentially boil down to “I do not understand words”.
We have a very solid and explicit definition of what conscious means:
https://www.merriam-webster.com/dictionary/conscious#medicalDictionary
The best neuroscientists in the world do not even fully understand the complicated intricacies of the workings of the human mind. We do understand 100% exactly how computers work. There is no comparison.
I haven't even argued that AI is concious, I have said that we do not understand consciousness enough to define it. You have completely missed the point. This is Dunning-Kruger at its peak. You are vastly oversimplifying things, at what size of brain is an animal considered conscious or capable of thought? Define thought. We don't even have a solid understanding of life, much less thoughts. Yes we have a word definition, no we do not know where the line begins and ends. It is a philosophical question that remains unanswered.
This definition is already making so many assumptions and over simplifications. https://www.merriam-webster.com/dictionary/thought
We understand that the brain is what gives us our cognition, we understand that the brain is made up of chemical and electrical signals. We are not special.
This dude is a Dunning-Kruger case study. Simply by the shear overconfidence
I’m sorry that you have trouble grasping so many basic concepts.
Ironic of you to reference Dunning-Kruger.
All animals are conscious as far as I’m aware, though the complexity of their thoughts may vary. They all act of their own volition, computers do not and cannot act of their own volition.
You do not have to fully understand every aspect of consciousness in order to recognize that inanimate objects do not have it. “Will” is a component of consciousness, this is commonly understood. Computers are tools, they have no will. Computers calculate, they do not think, these are different operations.
I am not oversimplifying things, you are under-thinking things that are already simple (likely due to cognitive offloading).
The argument basically: "'computers are not sentient, because I say that they are not sentient"
Don't forget the dictionary quotes and the personal attacks.
No, my argument is that your conceptualization of sentience is incomplete.
As if your comprehension, yet your arrogance is stunning because it is massive and is only dwarfed by your propensity for being wrong.
So your response boils down to "no u," and insulting. I believe we are done here.
For what it's worth, I thought it was an interesting read until the other person went full ad hominem.
If all you got from that explanation was “no u” then you need to work on your reading comprehension.
I agree that this conversation doesn’t seem to be bearing any fruit tho
Discussing the consciousness of neural networks is NOT being pro-AI. I'd say it's even the opposite, as it displays the danger of such a technology.
Most of the words in the definition have multiple interpretations, it is not as easy as you wish it would be.
Finally, we do NOT at all know how trained neural networks work. We know how training and inference work. We can read the neurons, but the shear size of weights make it a rather impenetrable black box we have to study, partially, to understand the goals of the emergent sub networks.
Fearmongering about the idea of a neural network gaining consciousness is basically like doing free advertising for AI companies lmao
You are ascribing a power to them that they do not possess
For the record, I think neural networks have a lot of useful applications; for example, radiological image recognition. The term “AI” basically has no meaning though, and is most commonly used to describe LLMs, a technology that has very few (if any) good real world applications.
I am not fearmongering the fact it helps or not AI companies is not relevant here.
You are getting off topic.