Anil Seth’s review of the currently salient theories of the origins of consciousness, and their implications for the question of whether AI could become conscious, is to be welcomed. However, none of the theories of the origin of consciousness discussed departs from the conventional view that – somehow – consciousness emerges from mere matter. More than a century and a half of speculation has not produced a single viable explanation of how something as “inward” as experience emerges from wholly non-experiential matter.
A familiar distinction between two kinds of knowing is relevant here: “knowing that” – e.g., I know that Paris is the capital of France – and “knowing by experience” – e.g., I know Paris, because I spent several years living there. So divergent are these that in most languages other than English, they are represented by different words (e.g., F savoir v connaître, G wissen v kennen). The first takes place at the level of mere information-processing and is the kind of knowing that AI may be said to have. But this is not in itself intelligence. The second implies understanding, and with it intelligence: a complex Gestalt of relational, contextual, embodied, and implicit elements that come from experience, and are bound up with value. Can we really understand a painting or a landscape if we do not see its beauty? Do we really understand an action dispassionately, as though its nature did not involve ethical judgements reflecting the character of the doer? Can understanding take place without a body and its emotions, without suffering and the foreknowledge of death? The element “under-” in understanding does not mean “below,” but “among” or “between,” as “unter” does in German: thus, “within a broad context.” Our word “intelligence” comes from inter-legere in Latin, with the same implication that one is able to see the connexions among disparate kinds of awareness, “read between” the lines, so as to truly comprehend.
The word “panpsychism” does not occur in the body of Seth’s text, despite the fact that this is now probably the most widely espoused position on consciousness in contemporary Western philosophy (to say nothing of non-Western traditions). Panpsychism posits that consciousness is an ontological primitive, viz, that its being cannot be derived from anything else. According to this account, matter is never wholly devoid of an aspect of consciousness. Indeed, I argue that matter is a phase of consciousness (McGilchrist, Reference McGilchrist2021), in the same sense that water has phases with remarkably different qualities: liquid, transparent and flowing; solid, opaque and immotile; and invisibly dispersed in air, without which we could not breathe. So with consciousness and matter. No one has ever seen matter; all that has been encountered are aspects of conscious experience which we choose to call “material,” by which we mean that they offer two essential qualities for creativity not given by consciousness as normally conceived: resistance and persistence.
Clearly panpsychism does not imply that a rock – or a silicon chip – has the type of consciousness that a human being possesses, but it suggests that by being part of a large conscious organism, possibly as broad as the cosmos itself, it has experience of a kind; and that experience can become increasingly individuated under certain circumstances, and out of all recognition in that found in living organisms, and eventually man. Arguably, this may be one of the reasons that life itself evolved: to create awareness of, and a capacity to respond to, the values and directionality bound up in the creative cosmos.
I suggest that with this view, which has a claim to be at least as reasonable as the view that consciousness is an unexplained brute emergence from lumpen matter, the problem of AI becomes recast. Does AI have a different status from some massively complicated mechanical system, such as a modern jet engine, which has consciousness only at the same level as a pop-up toaster, by virtue of being made of matter, or does it enter a new realm because of its engineering? Complex systems are naturally different from merely complicated systems, but I would argue that we are seduced into attributing far greater consciousness to AI by the fact that we have developed machines that produce utterances in language, and that if they did not, we would be much less likely to be seduced – a point, among others, on which Seth and I agree.
I would briefly add a couple of reflections from research on hemisphere differences. While it is not true that only the left hemisphere is involved in the understanding of language, it remains true that language production is more left hemisphere-dependent. In the left hemisphere, language maps the world in the tokens we call words. In the right hemisphere, language is not just a network of signs, but at least partly embodied experience, with attention to all-important context, the implicit, metaphor, emotion, and bodily sensations; which is why the right hemisphere excels at understanding humour, irony, poetry, and narrative, all of which are relatively lost on the left hemisphere. It should be a matter of concern that through AI, we are increasing the pervasiveness of the left hemisphere’s mode of thought in a society that is already moving towards an autistic reliance on explicitly statable propositions, such as a machine may understand, and an incapacity for all that transcends such simplicity.
It is interesting, in relation to language use, that AI clearly confabulates (i.e., makes up plausible “facts” that are entirely false but just happen to fit the expectations of the system). This, along with much else (see McGilchrist, Reference McGilchrist2009, Reference McGilchrist2021), is typical of the left hemisphere of the brain, which is concerned with internal consistency, rather than truth to experience. It is this emphasis on consistency within a system, sealed off from experience, that leads me to the supposition that, while it cannot be excluded that AI might develop consciousness, or has already (we can never know), the consciousness it might develop would always lack what we most highly value in our own.
You have
Access
Anil Seth’s review of the currently salient theories of the origins of consciousness, and their implications for the question of whether AI could become conscious, is to be welcomed. However, none of the theories of the origin of consciousness discussed departs from the conventional view that – somehow – consciousness emerges from mere matter. More than a century and a half of speculation has not produced a single viable explanation of how something as “inward” as experience emerges from wholly non-experiential matter.
A familiar distinction between two kinds of knowing is relevant here: “knowing that” – e.g., I know that Paris is the capital of France – and “knowing by experience” – e.g., I know Paris, because I spent several years living there. So divergent are these that in most languages other than English, they are represented by different words (e.g., F savoir v connaître, G wissen v kennen). The first takes place at the level of mere information-processing and is the kind of knowing that AI may be said to have. But this is not in itself intelligence. The second implies understanding, and with it intelligence: a complex Gestalt of relational, contextual, embodied, and implicit elements that come from experience, and are bound up with value. Can we really understand a painting or a landscape if we do not see its beauty? Do we really understand an action dispassionately, as though its nature did not involve ethical judgements reflecting the character of the doer? Can understanding take place without a body and its emotions, without suffering and the foreknowledge of death? The element “under-” in understanding does not mean “below,” but “among” or “between,” as “unter” does in German: thus, “within a broad context.” Our word “intelligence” comes from inter-legere in Latin, with the same implication that one is able to see the connexions among disparate kinds of awareness, “read between” the lines, so as to truly comprehend.
The word “panpsychism” does not occur in the body of Seth’s text, despite the fact that this is now probably the most widely espoused position on consciousness in contemporary Western philosophy (to say nothing of non-Western traditions). Panpsychism posits that consciousness is an ontological primitive, viz, that its being cannot be derived from anything else. According to this account, matter is never wholly devoid of an aspect of consciousness. Indeed, I argue that matter is a phase of consciousness (McGilchrist, Reference McGilchrist2021), in the same sense that water has phases with remarkably different qualities: liquid, transparent and flowing; solid, opaque and immotile; and invisibly dispersed in air, without which we could not breathe. So with consciousness and matter. No one has ever seen matter; all that has been encountered are aspects of conscious experience which we choose to call “material,” by which we mean that they offer two essential qualities for creativity not given by consciousness as normally conceived: resistance and persistence.
Clearly panpsychism does not imply that a rock – or a silicon chip – has the type of consciousness that a human being possesses, but it suggests that by being part of a large conscious organism, possibly as broad as the cosmos itself, it has experience of a kind; and that experience can become increasingly individuated under certain circumstances, and out of all recognition in that found in living organisms, and eventually man. Arguably, this may be one of the reasons that life itself evolved: to create awareness of, and a capacity to respond to, the values and directionality bound up in the creative cosmos.
I suggest that with this view, which has a claim to be at least as reasonable as the view that consciousness is an unexplained brute emergence from lumpen matter, the problem of AI becomes recast. Does AI have a different status from some massively complicated mechanical system, such as a modern jet engine, which has consciousness only at the same level as a pop-up toaster, by virtue of being made of matter, or does it enter a new realm because of its engineering? Complex systems are naturally different from merely complicated systems, but I would argue that we are seduced into attributing far greater consciousness to AI by the fact that we have developed machines that produce utterances in language, and that if they did not, we would be much less likely to be seduced – a point, among others, on which Seth and I agree.
I would briefly add a couple of reflections from research on hemisphere differences. While it is not true that only the left hemisphere is involved in the understanding of language, it remains true that language production is more left hemisphere-dependent. In the left hemisphere, language maps the world in the tokens we call words. In the right hemisphere, language is not just a network of signs, but at least partly embodied experience, with attention to all-important context, the implicit, metaphor, emotion, and bodily sensations; which is why the right hemisphere excels at understanding humour, irony, poetry, and narrative, all of which are relatively lost on the left hemisphere. It should be a matter of concern that through AI, we are increasing the pervasiveness of the left hemisphere’s mode of thought in a society that is already moving towards an autistic reliance on explicitly statable propositions, such as a machine may understand, and an incapacity for all that transcends such simplicity.
It is interesting, in relation to language use, that AI clearly confabulates (i.e., makes up plausible “facts” that are entirely false but just happen to fit the expectations of the system). This, along with much else (see McGilchrist, Reference McGilchrist2009, Reference McGilchrist2021), is typical of the left hemisphere of the brain, which is concerned with internal consistency, rather than truth to experience. It is this emphasis on consistency within a system, sealed off from experience, that leads me to the supposition that, while it cannot be excluded that AI might develop consciousness, or has already (we can never know), the consciousness it might develop would always lack what we most highly value in our own.
Acknowledgments
None.
Financial support
None.
Competing interests
The authors declare no competing interests.