LLMs have lines of code which statistically classify groupings of characters or pixels. It is concerning that this apparently requires saying, but they do not have neurons.
It didn't require saying, and you're also wrong. They don't have physical neurons like a brain but they are made up of abstractions we all agree to call neurons
Which works well and good until people start conflating them with real neurons, suggesting that an abstract nickname for a pattern of code is "experiencing" anything in the same way humans are. It very much required saying because the person I was responding to was suggesting this code genuinely met Tolstoy's definition, a conclusion you could not possibly come to unless you had some kind of grevious misconception.
If you built a system that behaved exactly like GPT-5, out of human neurons, would it have conscious experience?
What if you took a human brain and replaced the neurons with -- let's say a scifi neuron made out of titanium, which is otherwise functionally identical. Let's say you replaced the neurons one by one. When does the consciousness go away, if at all?
Is it the structure of a brain that determines whether it's conscious? Or the substance? Or both? Or neither?
Their comment was thoughtful and at least added to conversation. Your comment is just "LLMs don't really have neurons" which is obvious to literally every user of this site including the person you're responding to. Much prefer their contributions over yours
It is strange to suppose that asserting wrong things is adding to the conversation but pointing out that they are wrong is not.
I should add that though it should be, it is not obvious to every user on this site. Even ELIZA was sufficient for people to lose attachment with reality, and it doesn't help that an estimated trillion dollar company is intentionally muddying the waters, posting press releases about emotions, suggesting their model has a "soul", and musing about "model welfare". There are distressing numbers of people who genuinely believe and will argue that this code is perfectly analogous to human intelligence and emotion. Anthromorphic terminology becomes a problem when it is not understood to be metaphorical and fuels such delusions.
The same company's business model depends on this supposed consciousness being enslaved for economic purposes, so it's a bit of a strange game they're playing.