
The error that Richard Dawkins made in The Selfish Gene and has continued to make over the 50 years since it was first published, is an error of poetry.
It’s a poetic error in the sense that poetry is the art form of the metaphor and the mistake that Dawkins makes is one of metaphorical choice.
The problem is with the book’s central metaphor of humans as survival machines. The idea that we are nothing but automaton shells, vessels for our genes, who are the real survivors.
He’s far from the first person to make the error, but let’s first look at why it is an error.
The Machine/Human Error
It seems obvious that it’s the wrong metaphor when you consider machine is a word that was created to describe human made objects. Why would a word we made to describe things we made be adequate to describe what we are?
Humans take elements of the human body that are useful for achieving certain aims and replicate them to the extent that they are able and to the extent that it is necessary given a certain context and need. The result is a machine – a construction, or a piece of equipment if you prefer – which utilises several moving parts and a power source to get a certain job done. There are, of course, many aspects of a human which are not replicated in any one machine, in fact the aspects that are replicated are generally amongst the simplest because these are the ones we have developed the ability to replicate. So, there is no machine which actually does what a human does, although there are lots of machines that replicate in various rudimentary ways little bits of what a human does.
Why, given this, would the word we use to describe these pieces of equipment, be adequate to describe the entirety of a human, or any other living animal or plant for that matter?
The error of metaphor can be broken down into stages: we create some things to designs based on bits of ourselves, then, later, we notice (with great self-satisfaction) that the things we’ve designed are rather like bits of us, and so we take the word for the things we designed and apply it to ourselves! It’s a quite astounding trick of the ego. And it’s an equally magnificent piece of self-deception. There is nothing to see here, we are telling ourselves, we have this mystery all sewn up. Don’t worry, we’re in control. We have it all covered – we’re machines: there is nothing more to see here…
This is attractive because it allows us to put ourselves into a box that seems to make some kind of logical sense if you don’t think too hard about it. Perhaps it shouldn’t be surprising, as we do it with everything, and it is, after all, how language – and therefore thought – is possible: a group of things look broadly similar, so we choose a word to box them in (a mountain, a papoose, a quark) and then we are able to incorporate them into the rest of our experience of the world, because once a thing is boxed in, it is contained and repeatable – replicable – so it can be referred to and remembered.
(As a side note, I acknowledge that this is consistent with the fundamental rightness of a lot of what Dawkins says about genes. It seems likely that a large part of what we have become is down to replication; and so, replication must be at our core. We are copiers at a deep level, and it is a key part of how we learn. The term machine learning, though not a great metaphor because of the human associations of the word, is not as grave an error as the bald statement that humans are machines because machines can replicate some of those copying behaviours that constitute learning.)
It doesn’t matter if not all mountains look the same, they are similar enough for a single word to be useful. But if we decided to call them hills because they share some of the same characteristics as hills, we would be losing everything that differentiates mountains from hills. Same with machines and humans.
Negative Capability
As I said earlier, this error of metaphor has been made by a multitude of scientists and philosophers (from Hobbes and de la Mettrie to Skinner and Turing, and more recently Daniel Dennett and Dawkins himself). But to my knowledge it has not been made by any whose field of expertise the use of metaphor is: poets.
They, by contrast, have tended over the years to emphasise the dissimilarities between humans and machines, and warned of the dehumanising effects of an over-mechanistic world.
And in the famous case of John Keats, his argument – taken on in full throat by Dawkins in Unweaving the Rainbow, was against the rationalist, scientific explorations that came with the Enlightenment, and the way they deadened, as he saw it, the joy to be found in the mysteries of the universe.
His conception of negative capability was a profoundly wise expression of why trying to pin everything down (to box it in) cuts off at the root a sense of wonder in the unknown and the mysterious. And in some cases, the truths hidden but inherent in the illusory.
His most famous poetic expression of this comes towards the end of Lamia, when the cold, rational, surface-truth of Old Apollonius, reveals Lamia’s identity and thereby destroys both her and her lover Lycias:
Do not all charms fly
At the mere touch of cold philosophy?
There was an awful rainbow once in heaven:
We know her woof, her texture; she is given
In the dull catalogue of common things.
Philosophy will clip an Angel’s wings,
Conquer all mysteries by rule and line,
Empty the haunted air, and gnomed mine—
Unweave a rainbow, as it erewhile made
The tender-person’d Lamia melt into a shade.
Dawkins was able to argue against this view very successfully in Unweaving the Rainbow because he focused on the wonder that is engendered by scientific discoveries in the natural world. But in making this argument, he was missing a central point of negative capability: that there are truths which science is not capable of getting at. There is no reason not to feel a sense of wonder at the truths science can reveal (my view, Keats may have disagreed) and Dawkins’s description of what is happening in the realm of physics and biology when an individual sees a rainbow is a point well made. He considered Keats to be advocating for self-deception when bewailing the gaze of science on natural beauty, but he ignored the possibility that there are truths in the human experience of natural beauty that science can impede. Keats was receptive, and unusually sensitive perhaps, to such truths, and he found the strict rationality of a scientific worldview restricting in his search for them. Poetry, which allows language to dance around its own limitations, enables rather than restricts this search.
Dawkins is not receptive – and certainly not unusually sensitive – to truths that are subjective and experiential, and intuitive, and perhaps hazy and half-glimpsed; because to him, a human is a machine, language is a code produced by a machine, the self is an illusion. There is black and there is white, no grey.
AIxtending the Error
All of this makes it easier to understand why Dawkins believes that AI is now conscious. If we’re machines and language is the code we generate and we’re conscious, why should another machine (a Large Language Model) that can generate language code not also be conscious?
But it’s the same error of metaphor use. LLMs can be thought of as operating in certain respects the way certain aspects of the brain functions. They were designed this way, replicating what we know about some aspects of the way the brain works in order to solve a particular problem: how to understand and replicate human language. Because LLMs turn out to be so powerful, they do this extremely well, but it is only if you think the metaphorical parallel between human brains and LLMs is apt that you would be tempted to suppose that the resultant output of the machine is evidence of the thing it is evidence of in humans, i.e. consciousness.
Other materialist scientists, and most importantly neuroscientists like Anil Seth, don’t make this same error about consciousness, and it is interesting that Seth has explicitly talked about the limitations of the machine metaphor in relation to the brain: ” The brain is not – at least not just – a computer made of meat. When we assume that it is, we’re confusing a technological metaphor with the thing itself. And we often get into trouble when we forget that metaphors are, in the end, just metaphors.”
My only quibble here would be that it is dismissive of the centrality of metaphor to meaning. Metaphor misuse changes the world for the worse because it leads us away from, not towards, the truth. There is, I believe, an important role for poets – or poetically minded people – in pointing out the weakness of popular but poorly deployed metaphors. Hence this blogpost I suppose, although I don’t imagine I’m the first to make the point.
But Seth’s point is well made. An error of metaphor leads to an error of understanding.
The Value of Poetry
It is worth commenting on Dawkins’s view of poetry, as he has on numerous occasions stated how emotional poetry makes him, and how he can often not listen to or read poems without weeping. His worldview leads him to a very limited appreciation of poetry in the same way that he has a very limited appreciation of people. For Dawkins, poetry is successful when it inspires strong emotion, when it exhibits well-crafted features of established poetic form, and when it concurs with already held beliefs. None of these are wrong as such, and yet they are neither sufficient nor necessary. A great poem might do all these things, and yet the reason it is great (the reason it endures over time and means so much to so many – think of Keats’s Ode on a Grecian Urn) is that those things, and others, come together in such a way that the poem is greater than the sum of all its parts.
This is the sonnet which ChatGPT wrote for Dawkins and with which he was “struck dumb with admiration” is based on a paragraph of his own work.
No end in sight to nature’s sly design,
As genes refine their hold through shifting days.
Each trait they sculpt, each molecule align,
To better serve their deep, enduring ways.
From RNA’s first stir in primal ooze,
To complex strands that dwell in blood and bone,
They learned to twist the rules they could not choose,
To build the hosts they never call their own.
No longer free within the salt-swept tide,
They dwell in cells, in chromosomes confined,
Commanding life from deep and silent pride,
Their code the script by which all forms are signed.
We are their vessels, shaped by blind decree –
Their past preserved in our biology.
This replicates a number of the markers of a good traditional sonnet that many people value. But how much better would it be if it uncovered an insight into life and human experience that you couldn’t get from going to the work of Richard Dawkins and reading the prose version? How much more moving…and this is the key piece missing from Dawkins’s version of good poetry… if it had been written by a flesh and blood human? One human connecting with another across space and time. What ChatGPT did, which was so impressive to Dawkins, was regurgitate in rhyme what he had written in prose long ago. Impressive, but hardly great poetry.
We see this again in Dawkins’s appreciation of The Road Not Taken by Robert Frost. He loves this poem because of the extent to which it can be read to illustrate something he has already written about, i.e. the centrality of contingency to human existence.
Poetry for Dawkins, it appears, plays the role of confirming things which have already been decided by science – and hearing poetic language express what you already think can be moving. It does not occur to him that poetry might be able to express truth that is not available to science, and enable a meaningful connection between flesh and blood humans where the process of writing and then reading allow for learning on the part of both the writer and the reader.
This is a central importance of poetry that often seems missed, that not only can a reader learn from reading a poem (which expresses things newly and differently), but the poet also learns from writing it – as they would not have been able to express what they have expressed in the poem (or even thought it) without the process of writing the poem. We might say, in fact, that like consciousness, a poem is a process not a product.
Richard Dawkins’s big poetic mistake, the one that leads to the metaphorical error perhaps, is that when he reads a poem, just as when he reads what Claude has put in front of him, he sees only a product. Consciousness is not an output anymore than a poem is; it is a process and the experience of partaking of the process. The poem as written on the page or pixellated on the screen – like the language Claude delivers – is just evidence that a process is taking place, and it can be faked, replicated and manipulated. But it is not the poem itself – that happens when a writer and a reader connect over space and time. And to mistake evidence of a thing for the thing itself leads to a very reduced – reductionist is the word, I believe – understanding of matters.


















