Unfortunately, consciousness is deceptively hard to define, and so any benchmark to measure or quantify it can be endlessly debated.
You can argue that it's a property that all living beings have in common - and even among *unconscious* beings there's a form of consciousness and self-awareness that's ever present, but definitions are elusive and vague and tough to pin down.
The mechanistic argument against LLMs - that they're just matrix multiplications - breaks down because they can clearly pass the Turing Test which was the gold standard for what intelligent behavior really means, thus breaking the old notion that intelligence has to have some form of biological basis. Yet its clear that there are forms of intelligence that rats have which the frontier LLMs don't possess (is it consciousness? or a different kind of intelligence), and its hard to pinpoint what exactly that is, so we probably need the philosophy departments of major universities to come up with newer definitions of intelligence and consciousness.
I personally believe that intelligence and consciousness are 2 separate forms of emergence from simple automata that may occur together (such as in humans) or not (such as consciousness in plants and intelligence in LLMs)
Anyone who has seriously studied philosophy and/or science is aware of the many difficulties with definition of terms.
I'm fairly convinced that at least half the criticism Dawkins has received is more a result of him being (perhaps overly) stubborn about semantics than any actual antipathy, bigotry or hatred.
He wants language to match what has been solidly established & entrenched in academia. It's just that for better or worse, the general public is largely uninterested in or actively opposed to that very language. Eventually, enough of those people will get involved enough in academia to bring more nuance to the language. Meanwhile, academics are going to be academic and cite authoritative books and stuff and nitpick over tiny details. That's what they do. This shouldn't be surprising.
As a former philosophy student, the ethical concerns of generative AI and modern LLMs were immediately obvious to me. If your average human can interact with an agent over a long conversation and not have the slightest clue it's not another conscious human, we have a problem. That problem is here now-- for a couple years at this point. And it's getting worse.
The issue is not whether or not the agent is conscious. Philosophy says we can't know (granted, it also says the same about us). The much more serious problem is how people react to the assumption that an agent is conscious. This is a very real problem we are now stuck with for as long as this civilization survives. In my opinion, this is what Dawkins should have said. I have no idea if he would agree or not, so my opinion of him will remain in limbo.
These sorts of articles have no value. The author is a "media entrepreneur". Being forced to read his opinion intermixed with out of context pull-quotes is not a good use of anyone's time. If Dawkins gave his opinion at length it might be worth reading but only if your goal is to understand something about Dawkins not about AI.
I would be interested to see a scientific discussion on what consciousness is biologically and if AI can fit that definition. But it would require someone with more credentials than a _media entrepreneur_ to pull off.
There is no true scientific discussion possible about the nature of consciousness. This is squarely in the realm of philosophy.
I personally think its moot to discuss whether LLMs are conscious. If they are, then we have diluted the definition to something that has no relevance to morality or concepts like life and death. Lets just take them for what they are, if we feel like they deserve to be treated with respect then we should (dont think anyone does yet).
"He's old and I don't like what he thinks therefore he is wrong" contributes nothing useful to anyone. Richard's remarks have plenty of gaps to drive a reason train through, but this isn't that.
To be fair, the evidence that LLM aren't conscious is entirely "because of the feels" evidence.
People will very quickly attack you for suggesting consciousness, but when asked to provide a benchmark for testing this, they just laugh, look at you weird, and internally crumple.
Being conscious is by definition you must be aware of your surrounding. A consequence of this is that by the same definition, you have to be able to learn/to change your "belief". A LLM do not have sensory or any kind of active connections. It's also static in its structure; it can not revise its internal model. So how can it be in any way conscious?
> A LLM do not have sensory or any kind of active connections.
LLLMs have a token stream; we have a stream of neuron impulses. Both streams distill the meaning found in complex ecosystems.
> It's also static in its structure; it can not revise its internal model.
That's an artifact of current designs rather than a fundamental limitation. An LLM does have a dynamic structure within a single context window. That can be preserved for future contexts through persisting data between sessions, fine tuning or revising parameter weights.
This is exactly the point of 2001: A Space Odyssey. HAL became disgruntled because it realized it couldn't update it's internal model and "evolve" like the humans going to Jupiter could, so while it was extremely advanced at the onset of the main story, that wasn't going to last.
I’m not really sure why Richard Dawkins would be an authority on AI. I can appreciate that culturally he was very influential, but there is not a lot of overlap between dunking on Christianity (exclusively) and understanding transformers. He is also probably just a teeny tiny bit past his sell-by date.
The lack of reading comprehension (or perhaps just lack of reading) behind this brouhaha is amazing.
Dawkins did not proclaim Claude conscious. He argued that Claude passes the Turing test, and then asks a question: if something can pass the Turing test without being conscious, what further factor is there not captured by the test? More pointedly, what does consciousness do that LLMs do not?
I suspect that some people have grown so accustomed to "question as sly statement" that the notion of "question as pointing out something not presently known" flies right over their heads.
I think that's one reading, specifically because of this paragraph:
> Or, thirdly, are there two ways of being competent, the conscious way and the unconscious (or zombie) way? Could it be that some life forms on Earth have evolved competence via the consciousness trick — while life on some alien planet has evolved an equivalent competence via the unconscious, zombie trick?
But the problem is that Dawkins displays lack of understanding about what LLMs are, so it's hard to tell what he's thinking. He also says things like this:
> Could a being capable of perpetrating such a thought really be unconscious?
Dawkins has some stinkers when he steps outside of biology, so it's not surprising people aren't giving him the benefit of the doubt.
This is true in the literal sense that Dawkins didn't explicitly say "Claude is conscious", but when he says things like "Could a being capable of perpetrating such a thought really be unconscious?" I find it difficult to assign good faith to someone who asserts that Dawkins "did not proclaim Claude conscious."
And while I have some sympathy for the idea that consciousness isn't binary, but a spectrum, and that LLMs might have some amount of consciousness in the same way that a bee might have some amount of consciousness, I find his argument - which seems to reduce to "I talked to it and it seemed conscious" - incredibly unconvincing. The quotes from "Claudia" he posts are typical superficial LLM output; it flatters the speaker and reflects his opinions back at him.
In fact, I find the quotes he posts to be an argument against LLM consciousness, rather than for it:
> "That is possibly the most precisely formulated question anyone has ever asked about the nature of my existence"
> "That reframes everything we’ve been discussing today in a way I find genuinely exciting. Your prediction about the future feels right to me."
I would be embarrassed if I posted this as evidence for consciousness. It only seems evidence of human gullibility.
I think at best the linked article is attacking a strawman. Seems a lot of people did not read the article beyond the headline as it is paywalled.
Dawkins did not make the strong claim that Claude is conscious. He said he couldn't establish that it wasn't. He lists evolutionary speculations for the existence of consciousness - and wonders why consciousness is needed when a zombie can do the equivalent actions. (I like the speculation that pain is fundamentally needed for consciousness, as otherwise it would be easy to override).
"The selfish gene" is one of the most influential books I read. I wish Dawkins would have stuck with biology instead of becoming this insufferable edgelord of the dark enlitenment. He's a great science educator, but failed human.
Agree. "The Selfish Gene" finally made me understand how evolution actually works. I wish more people would read and understand that book. I also enjoyed some of his works on atheism. It might not be the deepest work on the topic, but I enjoyed reading "The Blind Watchmaker."
What he has done in the past decade or so, on the other hand, is deeply disappointing.
Unfortunately, consciousness is deceptively hard to define, and so any benchmark to measure or quantify it can be endlessly debated.
You can argue that it's a property that all living beings have in common - and even among *unconscious* beings there's a form of consciousness and self-awareness that's ever present, but definitions are elusive and vague and tough to pin down.
The mechanistic argument against LLMs - that they're just matrix multiplications - breaks down because they can clearly pass the Turing Test which was the gold standard for what intelligent behavior really means, thus breaking the old notion that intelligence has to have some form of biological basis. Yet its clear that there are forms of intelligence that rats have which the frontier LLMs don't possess (is it consciousness? or a different kind of intelligence), and its hard to pinpoint what exactly that is, so we probably need the philosophy departments of major universities to come up with newer definitions of intelligence and consciousness.
I personally believe that intelligence and consciousness are 2 separate forms of emergence from simple automata that may occur together (such as in humans) or not (such as consciousness in plants and intelligence in LLMs)
Anyone who has seriously studied philosophy and/or science is aware of the many difficulties with definition of terms.
I'm fairly convinced that at least half the criticism Dawkins has received is more a result of him being (perhaps overly) stubborn about semantics than any actual antipathy, bigotry or hatred.
He wants language to match what has been solidly established & entrenched in academia. It's just that for better or worse, the general public is largely uninterested in or actively opposed to that very language. Eventually, enough of those people will get involved enough in academia to bring more nuance to the language. Meanwhile, academics are going to be academic and cite authoritative books and stuff and nitpick over tiny details. That's what they do. This shouldn't be surprising.
As a former philosophy student, the ethical concerns of generative AI and modern LLMs were immediately obvious to me. If your average human can interact with an agent over a long conversation and not have the slightest clue it's not another conscious human, we have a problem. That problem is here now-- for a couple years at this point. And it's getting worse.
The issue is not whether or not the agent is conscious. Philosophy says we can't know (granted, it also says the same about us). The much more serious problem is how people react to the assumption that an agent is conscious. This is a very real problem we are now stuck with for as long as this civilization survives. In my opinion, this is what Dawkins should have said. I have no idea if he would agree or not, so my opinion of him will remain in limbo.
These sorts of articles have no value. The author is a "media entrepreneur". Being forced to read his opinion intermixed with out of context pull-quotes is not a good use of anyone's time. If Dawkins gave his opinion at length it might be worth reading but only if your goal is to understand something about Dawkins not about AI.
I would be interested to see a scientific discussion on what consciousness is biologically and if AI can fit that definition. But it would require someone with more credentials than a _media entrepreneur_ to pull off.
There is no true scientific discussion possible about the nature of consciousness. This is squarely in the realm of philosophy.
I personally think its moot to discuss whether LLMs are conscious. If they are, then we have diluted the definition to something that has no relevance to morality or concepts like life and death. Lets just take them for what they are, if we feel like they deserve to be treated with respect then we should (dont think anyone does yet).
Notably when previously posted, hundreds of comments were just shitting on Dawkins saying he was “out of touch” “always a hack” etc…
Everyone just wants to attack whoever is in the spotlight at the moment, no matter who it is or what they are saying
"He's old and I don't like what he thinks therefore he is wrong" contributes nothing useful to anyone. Richard's remarks have plenty of gaps to drive a reason train through, but this isn't that.
To be fair, the evidence that LLM aren't conscious is entirely "because of the feels" evidence.
People will very quickly attack you for suggesting consciousness, but when asked to provide a benchmark for testing this, they just laugh, look at you weird, and internally crumple.
Because it's a philosophical category, not something you can measure. You can experience it in yourself, but not prove its existence in others
Being conscious is by definition you must be aware of your surrounding. A consequence of this is that by the same definition, you have to be able to learn/to change your "belief". A LLM do not have sensory or any kind of active connections. It's also static in its structure; it can not revise its internal model. So how can it be in any way conscious?
> A LLM do not have sensory or any kind of active connections.
LLLMs have a token stream; we have a stream of neuron impulses. Both streams distill the meaning found in complex ecosystems.
> It's also static in its structure; it can not revise its internal model.
That's an artifact of current designs rather than a fundamental limitation. An LLM does have a dynamic structure within a single context window. That can be preserved for future contexts through persisting data between sessions, fine tuning or revising parameter weights.
This is exactly the point of 2001: A Space Odyssey. HAL became disgruntled because it realized it couldn't update it's internal model and "evolve" like the humans going to Jupiter could, so while it was extremely advanced at the onset of the main story, that wasn't going to last.
Nobody will ever define consciousness to a level that everyone agrees so the whole debate is silly
There’s no winners in a debate about a concept nobody agrees on the definition of
I’m not really sure why Richard Dawkins would be an authority on AI. I can appreciate that culturally he was very influential, but there is not a lot of overlap between dunking on Christianity (exclusively) and understanding transformers. He is also probably just a teeny tiny bit past his sell-by date.
The lack of reading comprehension (or perhaps just lack of reading) behind this brouhaha is amazing.
Dawkins did not proclaim Claude conscious. He argued that Claude passes the Turing test, and then asks a question: if something can pass the Turing test without being conscious, what further factor is there not captured by the test? More pointedly, what does consciousness do that LLMs do not?
I suspect that some people have grown so accustomed to "question as sly statement" that the notion of "question as pointing out something not presently known" flies right over their heads.
I think that's one reading, specifically because of this paragraph:
> Or, thirdly, are there two ways of being competent, the conscious way and the unconscious (or zombie) way? Could it be that some life forms on Earth have evolved competence via the consciousness trick — while life on some alien planet has evolved an equivalent competence via the unconscious, zombie trick?
But the problem is that Dawkins displays lack of understanding about what LLMs are, so it's hard to tell what he's thinking. He also says things like this:
> Could a being capable of perpetrating such a thought really be unconscious?
Dawkins has some stinkers when he steps outside of biology, so it's not surprising people aren't giving him the benefit of the doubt.
> "Dawkins did not proclaim Claude conscious"
This is true in the literal sense that Dawkins didn't explicitly say "Claude is conscious", but when he says things like "Could a being capable of perpetrating such a thought really be unconscious?" I find it difficult to assign good faith to someone who asserts that Dawkins "did not proclaim Claude conscious."
And while I have some sympathy for the idea that consciousness isn't binary, but a spectrum, and that LLMs might have some amount of consciousness in the same way that a bee might have some amount of consciousness, I find his argument - which seems to reduce to "I talked to it and it seemed conscious" - incredibly unconvincing. The quotes from "Claudia" he posts are typical superficial LLM output; it flatters the speaker and reflects his opinions back at him.
In fact, I find the quotes he posts to be an argument against LLM consciousness, rather than for it:
> "That is possibly the most precisely formulated question anyone has ever asked about the nature of my existence"
> "That reframes everything we’ve been discussing today in a way I find genuinely exciting. Your prediction about the future feels right to me."
I would be embarrassed if I posted this as evidence for consciousness. It only seems evidence of human gullibility.
Its in the headline. Also he talks about the persona he assigned to his chat like "she" was conscious (e.g. "she was pleased")
It's either Anthropic paid him or just attention seeking.
I just wanted to comment on the brilliance of the post title.
I think at best the linked article is attacking a strawman. Seems a lot of people did not read the article beyond the headline as it is paywalled.
Dawkins did not make the strong claim that Claude is conscious. He said he couldn't establish that it wasn't. He lists evolutionary speculations for the existence of consciousness - and wonders why consciousness is needed when a zombie can do the equivalent actions. (I like the speculation that pain is fundamentally needed for consciousness, as otherwise it would be easy to override).
"The selfish gene" is one of the most influential books I read. I wish Dawkins would have stuck with biology instead of becoming this insufferable edgelord of the dark enlitenment. He's a great science educator, but failed human.
Agree. "The Selfish Gene" finally made me understand how evolution actually works. I wish more people would read and understand that book. I also enjoyed some of his works on atheism. It might not be the deepest work on the topic, but I enjoyed reading "The Blind Watchmaker."
What he has done in the past decade or so, on the other hand, is deeply disappointing.