Towards the beginning of his paper, Seth announces his assumption that “there is a fact-of-the-matter about whether something is conscious” and “being conscious is not determined by social or linguistic consensus.” He is far from alone in these convictions. They underlie almost every contribution to the contemporary philosophical debate around consciousness, which is no surprise because the metaphysical intuitions behind such pronouncements are deeply embedded in our conceptual make-up. To uproot those intuitions – to “show the fly the way out of fly-bottle,” as Wittgenstein has it, or to dispel our “primal confusion,” as Garfield frames the matter in a Buddhist context – is profoundly difficult (Wittgenstein, Reference Wittgenstein1953/2009, §309; Garfield, Reference Garfield2015, pp. 9–11; see also Shanahan, Reference Shanahan2010, ch.1).
First, let’s note that how we use the word “consciousness” is a matter of social and linguistic consensus. Such consensus is presently lacking in the context of AI, and we’ll come to that shortly, but let’s stick to more settled usage for now. Our first step is to set aside the question of what consciousness “is,” and to ask instead how words like “consciousness,” and its many relatives (such as “awareness,” “attention,” and so on), are used. The crucial step, though, is to see that, having spelt out how these words are used, there are no residual philosophical problems. The challenge here is not the spelling out, but the insight that there is no more to be said, from a metaphysical point of view.
There undoubtedly is more to be said from a scientific point of view. The relationship between our usage of such words, the human (and animal) behaviour such usage relates to, and the brain processes that underlie that behaviour, certainly merits scientific study. Moreover, the fruits of this study will feed back into the way we use the words in question, and more generally into the nexus of social practices in which the use of those words is embedded. In this way, an issue such as the welfare of farmed octopuses can become a matter of societal and legislative concern, as it recently has in the UK. However, there is no need to resort to philosophical language to characterise what is going on here. We don’t need to think procedures like this reveal what consciousness is, where the word “is” carries a metaphysical load.
Now let’s extend this approach to the topic of artificial intelligence (Shanahan, Reference Shanahan2024). The example of the octopus shows that the way we use words is subject to change, even philosophically difficult words like “consciousness” and its kin. The development of sophisticated artificial intelligence is putting our language under particular pressure, and there is no consensus yet on how it should adapt to these exotic mind-like artefacts. We are in a period of upheaval while scientists, scholars, and the public collectively figure out how to adapt and extend our ways of talking and thinking to accommodate them. Importantly, this process can only draw on what is public, what is manifest in our shared world, namely, behaviour and the mechanisms underlying that behaviour.
Let’s recap. The philosophical strategy here is to replace the metaphysically loaded question of what consciousness is with questions about how words like “consciousness” are used. When confronted with scenarios where usage is not settled, such as the arrival of powerful AI, our strategy is to outline the process whereby usage changes, through a society-wide conversation informed by scientific investigation. But now for the hard part, which is to see that there is nothing more to be said, no residual philosophical problem. This is a great challenge because it forces us to confront our deep-seated dualistic intuitions, our primal confusion.
Seth betrays his own dualistic intuitions by asserting that “[w]e humans know we’re conscious” and immediately issuing the caveat that “strictly, we know it for ourselves and infer it about others.” The feeling that consciousness is something – some “thing” – that we only truly “know” from our own case and merely infer of others, is almost irresistible. It is, according to intuition, the most intimate form of knowledge by acquaintance: immediate, privileged, incorrigible, but above all private. It has led innumerable philosophers astray, from Descartes in the 17th century to Nagel and Chalmers in the 20th, and it continues to do so today in the 21st.
Buddhist philosophers have long known how to deconstruct the metaphysical thinking behind dualism, at least since the time of Nāgārjuna in the 3rd century (Siderits & Katsura, Reference Siderits and Katsura2013; Garfield, Reference Garfield2015). In Western philosophy, potent antidotes to dualistic thinking can be found in the work of such 20th-century figures as Heidegger, Derrida, and Wittgenstein. The so-called private language remarks in Wittgenstein’s Philosophical Investigations are especially pertinent.
Reflecting on the temptation to say “it is only from my own case that I know what the word ‘pain’ means,” Wittgenstein invites us to imagine that “everyone had a box with something in it: we call it a ‘beetle’. No one can look into anyone else’s box, and everyone says he knows what a beetle is only by looking at his beetle” (Wittgenstein, Reference Wittgenstein1953/2009, §293). He points out that, for all anyone would know, the contents of each person’s box might be different. The contents could be constantly changing, or the box could even be empty. The private “thing” itself would play no part in our language: “it cancels out, whatever it is.” Hopefully, the analogy with consciousness in general is obvious.
So, where does this leave us with respect to biological naturalism and to Seth’s paper? While Seth frames his contribution as an argument concerning what consciousness “is,” it can equally be framed as an argument for how we should use words like “consciousness.” It is, that is to say, a contribution to the ongoing society-wide conversation about how the language of consciousness should accommodate sophisticated artificial intelligence. The paper’s contribution is no less valid for being reframed in this way. It does, perhaps, render it a little more vulnerable to further critique. But that would be the subject of different commentary.
You have
Access
Towards the beginning of his paper, Seth announces his assumption that “there is a fact-of-the-matter about whether something is conscious” and “being conscious is not determined by social or linguistic consensus.” He is far from alone in these convictions. They underlie almost every contribution to the contemporary philosophical debate around consciousness, which is no surprise because the metaphysical intuitions behind such pronouncements are deeply embedded in our conceptual make-up. To uproot those intuitions – to “show the fly the way out of fly-bottle,” as Wittgenstein has it, or to dispel our “primal confusion,” as Garfield frames the matter in a Buddhist context – is profoundly difficult (Wittgenstein, Reference Wittgenstein1953/2009, §309; Garfield, Reference Garfield2015, pp. 9–11; see also Shanahan, Reference Shanahan2010, ch.1).
First, let’s note that how we use the word “consciousness” is a matter of social and linguistic consensus. Such consensus is presently lacking in the context of AI, and we’ll come to that shortly, but let’s stick to more settled usage for now. Our first step is to set aside the question of what consciousness “is,” and to ask instead how words like “consciousness,” and its many relatives (such as “awareness,” “attention,” and so on), are used. The crucial step, though, is to see that, having spelt out how these words are used, there are no residual philosophical problems. The challenge here is not the spelling out, but the insight that there is no more to be said, from a metaphysical point of view.
There undoubtedly is more to be said from a scientific point of view. The relationship between our usage of such words, the human (and animal) behaviour such usage relates to, and the brain processes that underlie that behaviour, certainly merits scientific study. Moreover, the fruits of this study will feed back into the way we use the words in question, and more generally into the nexus of social practices in which the use of those words is embedded. In this way, an issue such as the welfare of farmed octopuses can become a matter of societal and legislative concern, as it recently has in the UK. However, there is no need to resort to philosophical language to characterise what is going on here. We don’t need to think procedures like this reveal what consciousness is, where the word “is” carries a metaphysical load.
Now let’s extend this approach to the topic of artificial intelligence (Shanahan, Reference Shanahan2024). The example of the octopus shows that the way we use words is subject to change, even philosophically difficult words like “consciousness” and its kin. The development of sophisticated artificial intelligence is putting our language under particular pressure, and there is no consensus yet on how it should adapt to these exotic mind-like artefacts. We are in a period of upheaval while scientists, scholars, and the public collectively figure out how to adapt and extend our ways of talking and thinking to accommodate them. Importantly, this process can only draw on what is public, what is manifest in our shared world, namely, behaviour and the mechanisms underlying that behaviour.
Let’s recap. The philosophical strategy here is to replace the metaphysically loaded question of what consciousness is with questions about how words like “consciousness” are used. When confronted with scenarios where usage is not settled, such as the arrival of powerful AI, our strategy is to outline the process whereby usage changes, through a society-wide conversation informed by scientific investigation. But now for the hard part, which is to see that there is nothing more to be said, no residual philosophical problem. This is a great challenge because it forces us to confront our deep-seated dualistic intuitions, our primal confusion.
Seth betrays his own dualistic intuitions by asserting that “[w]e humans know we’re conscious” and immediately issuing the caveat that “strictly, we know it for ourselves and infer it about others.” The feeling that consciousness is something – some “thing” – that we only truly “know” from our own case and merely infer of others, is almost irresistible. It is, according to intuition, the most intimate form of knowledge by acquaintance: immediate, privileged, incorrigible, but above all private. It has led innumerable philosophers astray, from Descartes in the 17th century to Nagel and Chalmers in the 20th, and it continues to do so today in the 21st.
Buddhist philosophers have long known how to deconstruct the metaphysical thinking behind dualism, at least since the time of Nāgārjuna in the 3rd century (Siderits & Katsura, Reference Siderits and Katsura2013; Garfield, Reference Garfield2015). In Western philosophy, potent antidotes to dualistic thinking can be found in the work of such 20th-century figures as Heidegger, Derrida, and Wittgenstein. The so-called private language remarks in Wittgenstein’s Philosophical Investigations are especially pertinent.
Reflecting on the temptation to say “it is only from my own case that I know what the word ‘pain’ means,” Wittgenstein invites us to imagine that “everyone had a box with something in it: we call it a ‘beetle’. No one can look into anyone else’s box, and everyone says he knows what a beetle is only by looking at his beetle” (Wittgenstein, Reference Wittgenstein1953/2009, §293). He points out that, for all anyone would know, the contents of each person’s box might be different. The contents could be constantly changing, or the box could even be empty. The private “thing” itself would play no part in our language: “it cancels out, whatever it is.” Hopefully, the analogy with consciousness in general is obvious.
So, where does this leave us with respect to biological naturalism and to Seth’s paper? While Seth frames his contribution as an argument concerning what consciousness “is,” it can equally be framed as an argument for how we should use words like “consciousness.” It is, that is to say, a contribution to the ongoing society-wide conversation about how the language of consciousness should accommodate sophisticated artificial intelligence. The paper’s contribution is no less valid for being reframed in this way. It does, perhaps, render it a little more vulnerable to further critique. But that would be the subject of different commentary.
Acknowledgements
Thanks to Kerry Shanahan for extensive discussion on the topic of this commentary.
Financial support
This research received no specific grant from any funding agency, commercial, or not-for-profit sectors.
Competing interests
The author is a part-time employee of Google, a developer and provider of large language model technology, and is also a shareholder in Alphabet, Google’s parent company.