They're trained on all content they could access produced by humans. The op is asking for the most probable string of token extracted from that content following the prompt.
How sure are you that they're synthesizing an opinion vs just grabbing some existing opinion ("humans value in-group belonging over truth" is not a new idea) and regurgitating it? That they all agree on this kind of suggests as much. Which of course means that they are, ironically... just following the crowd.
It's really easy to come up with an opinion-shaped sentence, that doesn't mean any actual reasoning has gone into it. It's just a pundit-shaped sentence, something someone might say at a party. It's a maximally bland "controversial" opinion about humanity.
Not really… their training datasets strongly skew towards the contemporary West; LLM training data isn’t representative of the contemporary global average, let alone the historical global average.
As well as the skew of the raw training data, RLHF further skews it—a lot of the beliefs of the average 2026 human being (who belongs to the global South) are likely unsafe/misinformation/etc when judged by the standards of the AI vendors.
Training itself doesn't directly produce synthesis, -- it creates reproduction. Generalization during the process can result in novel output, but that can go either way as nothing selects for insight vs hallucination.
Models are also trained on synthetic material (as opposed to your 'human history'), which may result in fixating novel concepts and not just reproduction. But there is no particular reason to expect any reliable insights from the way this is done outside of the domain of machine verifiable objective results.
For this kind of "your take on X" one shots-- the model is almost certainly regurgitating some bias from training.
LLMs are not conscious and cannot have "takes". They stochastically produce a simulacrum of a human's takes based on their training, which is not the same thing.
But then if you observe human behaviour at a macro level don't we do the same?
I think we like to pretend that we don't. But even if you think about it from a bio/physics point of view, all we are is just a bunch of biological wires and neurotransmitter packages.
I know there's the whole unproven quantum tubule nonsense but even that would be a physics effect that we could learn about, manipulate and extort just as we have with everything else we've learnt about our realities.
That suggests they can't make inferences or produce any original insights, which is wrong. They can absolutely have takes.
Even if it's a simulacrum, it's a simulacrum based on nearly everything that was ever written. No one who has ever lived can claim to know what these models have trained on.
Trained on a very small portion of what has been written, heavily biased by what is available to American companies to acquire. You could argue that LLMs are somewhat representative of their training data (with some consideration to the way that they are trained) but to say it’s anything like a representation of what humans think is nonsense.
There's nothing small about it, and I didn't say it's a representation of what humans think. I said it's the models take on humanity based on having been trained on a far larger amount of text than any human has ever read
The problem with asking these models stuff like "what do you think about humanity" is that, even if it can synthesize an insight about humanity, it's just as likely it's regurgitating some opinion it's already seen. Which, given that this "controversial idea" is hardly new, is likely what it's doing. It's like a maximally boring controversial opinion, wake me up when it starts endorsing the bicameral mind or something actually weird.
Maybe it can produce interesting insights. But it's like asking a pundit for a soundbite: what are the odds they aren't just shuffling together something that sounds appropriate to the question?
Oh you have a convenient list that separates the conditions for consciousness from the ones that just happen to be incidentally true of the substrate? Can I see it?
They're trained on all content they could access produced by humans. The op is asking for the most probable string of token extracted from that content following the prompt.
How sure are you that they're synthesizing an opinion vs just grabbing some existing opinion ("humans value in-group belonging over truth" is not a new idea) and regurgitating it? That they all agree on this kind of suggests as much. Which of course means that they are, ironically... just following the crowd.
It's really easy to come up with an opinion-shaped sentence, that doesn't mean any actual reasoning has gone into it. It's just a pundit-shaped sentence, something someone might say at a party. It's a maximally bland "controversial" opinion about humanity.
> They're trained on all of human history,
Not really… their training datasets strongly skew towards the contemporary West; LLM training data isn’t representative of the contemporary global average, let alone the historical global average.
As well as the skew of the raw training data, RLHF further skews it—a lot of the beliefs of the average 2026 human being (who belongs to the global South) are likely unsafe/misinformation/etc when judged by the standards of the AI vendors.
> They're trained on all of human history,
Training itself doesn't directly produce synthesis, -- it creates reproduction. Generalization during the process can result in novel output, but that can go either way as nothing selects for insight vs hallucination.
Models are also trained on synthetic material (as opposed to your 'human history'), which may result in fixating novel concepts and not just reproduction. But there is no particular reason to expect any reliable insights from the way this is done outside of the domain of machine verifiable objective results.
For this kind of "your take on X" one shots-- the model is almost certainly regurgitating some bias from training.
LLMs are not conscious and cannot have "takes". They stochastically produce a simulacrum of a human's takes based on their training, which is not the same thing.
But then if you observe human behaviour at a macro level don't we do the same?
I think we like to pretend that we don't. But even if you think about it from a bio/physics point of view, all we are is just a bunch of biological wires and neurotransmitter packages.
I know there's the whole unproven quantum tubule nonsense but even that would be a physics effect that we could learn about, manipulate and extort just as we have with everything else we've learnt about our realities.
That suggests they can't make inferences or produce any original insights, which is wrong. They can absolutely have takes.
Even if it's a simulacrum, it's a simulacrum based on nearly everything that was ever written. No one who has ever lived can claim to know what these models have trained on.
Trained on a very small portion of what has been written, heavily biased by what is available to American companies to acquire. You could argue that LLMs are somewhat representative of their training data (with some consideration to the way that they are trained) but to say it’s anything like a representation of what humans think is nonsense.
It is a simulacrum, the end.
There's nothing small about it, and I didn't say it's a representation of what humans think. I said it's the models take on humanity based on having been trained on a far larger amount of text than any human has ever read
> That suggests they can't make inferences or produce any original insights
No it doesn't suggest that. A random number generator can also produce original insight (with different efficiency and effectiveness)
> They can absolutely have takes.
Sure, if you're willing to redefine "take"
I'd suggest that a random number generator is a great place to go if you want to get an original number.
And an LLM is a great place to go if you want to get an original token?
The problem with asking these models stuff like "what do you think about humanity" is that, even if it can synthesize an insight about humanity, it's just as likely it's regurgitating some opinion it's already seen. Which, given that this "controversial idea" is hardly new, is likely what it's doing. It's like a maximally boring controversial opinion, wake me up when it starts endorsing the bicameral mind or something actually weird.
Maybe it can produce interesting insights. But it's like asking a pundit for a soundbite: what are the odds they aren't just shuffling together something that sounds appropriate to the question?
Aren’t human takes also a simulacrum of other human takes based on training?
How do you know they are not conscious?
How you know God doesn't exist ?
You can't prove negative; but it doesn't satisfy any conditions for a positive
Oh you have a convenient list that separates the conditions for consciousness from the ones that just happen to be incidentally true of the substrate? Can I see it?
Maybe it's similar to why I'm attracted to women but I'm not attracted to mannequins.