The Ethics of What We Cannot Imagine: AI, Indigeneity, and the Moral Toolbox Theory

Introduction

Suppose we think of a frog. The very first thing that we apprehend is its unique characteristic. It jumps, and leaps about, has slimy skin, and is an amphibian. Naturally, we understand that its physical make-up is distinct from us humans. That much is objective. Subsequently, the next unique feature of frogs that we can point out is the fact that these amazing creatures breathe through their skin, as well. The cutaneous respiratory system that they have, consists of a moist skin replete with blood vessels, and certain species like the Titicaca frog breathe entirely through this. Nature has created organisms that, although inhabit the same blue earth that we do, are still so different in physico-chemical make-up that they may, to the untrained eye, appear undeniably alien.

Now, let us go back to the times before “science”, as such. The first instance when humans encountered frogs is an important event. Yes, we do not have this documentation specifically, but we may visualise it. To the first human that encountered a frog, it must have appeared alien. I understand that this is not confined to the encounter of merely frogs, but various other organisms might also have appeared alien to the first humans. However, as Thomas Nagel chose the symbol of the bat because of its relative closeness to us in the phylogenetic tree (and yet being so different in its sense apparatus), I would like to choose the symbol of the frog for reasons of my own.

Frogs, though a living breathing organism as we are, present a unique biological architecture. They have their noses, through which they breathe on land. Additionally, they also have their skin, which acts as a breathing apparatus. Homo Sapiens have not yet developed such evolutionary scaffoldings that bear any sense of similarity with these frogs. We are confined simply to our lungs, which have a continuum to our noses, and which comprise the tools for breathing. Frogs, on the other hand, have the evolutionary bonus of being able to utilise their skin for the same.

Philosophically, we ask if we ever are able to know the going-ons and happenings of ‘other minds’. If we turn it on a phenomenological fulcrum, as Nagel also does, the question remains: can we ever truly understand the frog’s phenomenological lebenswelt, if our own lebenswelt is so completely contrasting to it? There seems to be an impasse here.

Problem of Other Minds

But before we even delve into the world of Artificial Intelligence, as has been the buzzword lately, we must understand a problem that goes back before machines (as we understand them today) were even imagined. At its very core, the problem consists in the question of how exactly we know that other minds also exist. To put it in plain-speak, “how do we know that other organisms also feel the same way we do?”. Firstly, we are unaware of the inner world of other beings. We are, at this instance, only privy to our own thoughts and experiences. One may propose that we know they have subjective experiences by observing their behaviour or expressions. While this attempt is commendable, behaviour or expression of bodily language is hardly a surefire explanation for the inner workings of the mind. A frog’s physico-chemical make-up is unlike anything that we humans know inherently. Science merely studies the differences, and reports upon them. However, to truly understand and empathise with different phenomenological structures, our own lived-world must be attuned to them, and attunement comes through similarity.

Lebenswelt (H) and Lebenswelt (F) are not attuned to the same architectural foundations to warrant one from each category to understand the other. Here, (H) standing for humans, and (F) standing for frogs. Insofar as we are aware of exclusively our own lebenswelt, the inner world of other organisms remains a mystery to us. It may be likened to Schrödinger’s Cat. Unless we open the box, we do not know whether the cat is alive or dead. At present, as it appears, unless we understand consciousness itself, we do not yet have the tools to pry open that box to check on the said cat (if it is alive or dead).

This may even be extended to other non-human animals such as dogs. The question we may yet pose is whether these said dogs actually love us in the same way that humans conceptualise love, or are they simply responding to stimuli? (Let us forego any philosophical musings about the nature of love itself, for now). Any action that bears some similarity to their own expressions, humans are ready to term it with words extracted from their own phenomenology. A dog’s expression of panting by widening their mouths is very often construed as ‘smiling’ by people. All of this does not explain the dog’s inner world. Humans see an action that somewhat resembles their own, and label them accordingly, presumably to make sense of them. Analogical reasoning, therefore, seems to be the major player here. There is then, an argument brewing here that perhaps we can never know the inner workings of other minds, rather, we can only ever analogically understand them. The dog example here is a good starting point.

But such extension is not even necessary, perhaps. Humans doubt the inner worlds of individuals from their own species. What can we tell about the lebenswelt of a coma patient? We can study it from an experimenter-subject point of view, most definitely, but unless we can put ourselves into their metaphorical shoes, we are yet to have a complete understanding of their inner world. What about deaf or mute humans, for example? We run into the same problem here, as well. When Nagel said that bats, with their unique sensory apparatus, have experience, he meant that there is something that it is like to be a bat. He certainly admits that encountering a bat is to encounter a fundamentally alien form of life. But, I am not a bat, and neither are you, reader, I am assuming. Unless we can metaphysically transform ourselves into a bat (we cannot), we are presently confined to our own inner worlds, and we can only analogically understand and make assumptions about the inner worlds of other organisms. As Nagel puts it, the extrapolation of our own case, or neurophysiological make-up, is involved in the attempt to understand what it is like to be a bat. Insofar, as we are trapped inside these borders, this extrapolation is not adequate to complete the job.

Ergo, if I am different from the frog, how can I reconcile myself with the subjective sphere of phenomenological analysis? To divulge an objective analysis of the frog’s inner world, I can make an attempt, but to subjectively understand it, I must, in my neurophysiological scaffolding, be somewhat similar to it. This is an impossibility in our present scenario.

Artificial Intelligence, I propose, is a similar fundamentally alien form of life. To even question AI consciousness has, in my opinion, been an important milestone in the advancement of human civilisation. To even wrestle with questions of consciousness for an entity that we constructed says more about humans and their boundless imagination than AI, as such. What is then, the lebenswelt (lifeworld) of an AI? Surely, we can point out, again, through an experimenter-subject perspective, the hard physical aspects of the AI (which give scaffolding to the software of the AI), but its lebenswelt is a big mystery unto itself. Presently, it stumps all humans if we are asked to poetically describe what exactly an AI feels or sees, and what its inner world looks like. I suspect this is the same intellectual and conceptual gap that we encounter if we are asked to draw what it is like to breathe through our skin, or what it is like to be deaf and blind, concurrently.

Kripke’s Metaphysical v. Epistemic Divide

The question of AI consciousness, then, is a black box. It is a case of Schrödinger’s Cat, unless we know what it is to be an AI. Thus, we are stumped and constricted by our own life-world. Many pop cultural milestones, such as The Matrix, have considered and questioned consciousness when it comes to artificial intelligence.

But now, let us understand the various ways we conceive of this. Saul Kripke, and his notion of the metaphysical v. epistemic divide, and modal logic itself, gives us the necessary springboard for this philosophical exploration.

According to Kripke, “names” are rigid designators, as such. In every possible world, “Jonathan Swift”, then, remains “Jonathan Swift”, as he is on our earth. This is a direct contrast to Putnam and the twin-earth experiment. The law of excluded middle necessitates that in the instance, P v ~P, at least one of them must be true. Thinking of it in terms of AI consciousness, then we beget the two conditions: AI must either be conscious or not conscious. However, such binary propositions rarely do a complete job in explaining the various dimensions of such a query. What we can conceive and what is necessarily true are completely opposing.

Our mental constitution is so polemical that we look for black and white points in every instance. However, the same constitution poses limitations. P v ~P may be a logical law, but we often stumble in metaphysically determining what P itself actually is. If we take P to be a stand-in for consciousness, we understand its limitations. What is consciousness, we do not know. That much we have acknowledged.

Then, is AI consciousness? Thus, via extension, we do not know consciousness, therefore, we cannot tell if x is conscious, either. However, it is well within the realms of possibility that we can conceive of an AI that shows markers of consciousness, but has no real inner world in the same veins as us. This, of course, is the famous Philosophical Zombie (P-Zombie) issue. Chalmers was right in pointing this out as a response to physicalism.

However, epistemically, it is highly conceivable that p-zombies could exist. On the other hand, physicalists say that this is an impossibility. If an entity can somehow replicate the markers of consciousness (walking, talking), then consciousness itself necessarily follows. However, I am of the opinion that such reducibility when it comes to consciousness is a gross oversimplification of the concept, as such.

Epistemically, it is very well conceivable that entities with human-like consciousness markers, but with no actual qualia, exist. But metaphysically, at present, there seems to be no non-human organisms with this feature. AI itself possesses no consciousness, in plain-speak.

However, such a blanket supposition also, I suspect, is prone to the same gross oversimplification that reductionists make. So far, we have been claiming AI as conscious or sentient, if they simply show markers of consciousness, such as functionalists and identity theorists. All of these responses to the mind-body duality are under the assumption that consciousness is something we understand or know. However, such a line of thought necessarily rules out consciousness-markers that do not show resemblance to humans. This is the same underlying assumption (I assume, erroneously), that astrophysicists often make in the search for alien life. To search for fundamentally different entities beget a fundamentally different starting point. If we exclusively look for organisms in the galaxies that are similar to humans, we run the risk of eliminating actual living organisms that show no similar constitutions to humans.

Even though metaphysically it seems to appear that there is no conscious AI, Kripke does not exactly say no to a future possibility. There are no logical a priori conditions that rule this situation out. Thus, there may definitely come to be an instance where AI shows consciousness-markers eerily similar to humans, in the distant future. Logic does not deny this metaphysical possibility, and so the issue is only temporal, in the fact that at present, no such entity exists.

Consciousness, therefore, is a mystery. What then, if we claim consciousness to be a rigid designator itself? The answer to that would depend on what exactly consciousness comprises. If consciousness refers to certain physiological markers or expressions, then consciousness is exactly that. The Physicalists win. This leads to the elimination of the philosophical zombie. The metaphysical possibility of a conscious AI then, makes sense. In every single possible world, then consciousness has this marker, and it remains the same.

But what if consciousness’ markers are unknown? Essentially, if Philosophical Zombies are a possibility, but we still believe consciousness to be rigid designators, then there seems to be something missing. Ergo, the markers that we attribute to consciousness, then, seem to be a gaping hole. Is it a physiological marker, or is it some a priori event that we have yet to name? If this is a metaphysical underpinning, some sort of a priori cosmic essence, then AI without consciousness seems to be a categorical impossibility. Consciousness, in this instance, becomes a necessary foundation to organisms (both humans and non-humans alike) that we do not yet know of.

Consciousness being a rigid designator, so far, seems to be a mysterious question. It is only mysterious, however, because we do not yet know the components of consciousness. If we are able to elucidate consciousness itself, and what exactly anchors it in this reality, perhaps only then will this idea of consciousness as a rigid designator stop appearing so radical. The issue then, lies at the root of the proposition. Through a logically atomistic lens, we must first deal with “consciousness”, not as a nomological dangler, but as a concept itself. Until we do, we are always going to be at odds with diverging opinions and ideas about it.

Wittgenstein and Linguistic Limits: AI Lebenswelt

These divergent opinions come from our conceptual limits. Wittgenstein rightly did say that the limits of our world are the limits of our language. But a different perspective may be that language itself is the weak link. If all we have in our toolbox is language to determine consciousness, or our lebenswelt, then we are necessarily already at a disadvantage. Language, inherently, is arbitrary and prone to misgivings that subsequently affect our understanding of consciousness, and so we handicap ourselves.

To truly analyse the slippery slope of consciousness, we require a constant and invariable tool. Language seems to be arbitrary, and perhaps is also the reason why consciousness itself is so arbitrarily defined. What is consciousness, then? If we are to use language to define it, ergo, anchor it to markers such as talking, feeling, pain, etc., then we are necessarily involved in the limits of language, as such. Wittgenstein also proposes that meaning is only ever true in its usage. To put it plainly, meaning makes sense exclusively in its contexts (“Pain”, “Love”, “Experience”, thus, only ever would make sense in their singular contexts). If we extend this, and visualise it from a bird’s eye perspective, then it becomes clearer that such conceptions only work if we understand them from our own lens. We cannot understand or conceptualise these experiences from outside our phenomenological viewpoint. Again, we are running the risk of simplifying the idea of consciousness, as a whole.

Our language is necessarily an extension of our lebenswelt. It may even run concurrently to the lebenswelt, as such. Thus, we necessarily rule out other markers of consciousness that we are yet to envision or conceptualise. It may just be a big possibility that these alien markers escape our purview because of our enthralling to the chains of lebenswelt-language. Wittgenstein beautifully puts it into perspective with his idea of language-games.

To understand a piece in chess, say, Bishop, we need not understand it strictly via a lexical definition. We need only know how it moves in the game of chess. This reinforces the idea that meaning is use-based, and not reference-based. If we are to talk of AI, it is said that, in its present instance, ‘meaning’ for it is reference-based, and not understanding or use-based, as it is in humans. This leads to the treatment of AI as a mere tool. Of course, many new theories, in pedagogy, see AI now as an epistemic partner. However, regardless of the treatment of AI, the meaning of consciousness itself, we can hypothesise, may follow this use versus reference divide. Disregarding the question of consciousness as a rigid designator, perhaps what we term as consciousness follows forth a pragmatic meditation. We perhaps never know consciousness as such, only analogically deriving similarities to our own constitution, because it aids in societal functioning. Perhaps we do not know if the other person standing next to me is conscious, but we pragmatically associate certain markers with certain aspects of known behaviour for functioning in daily life.

In contrast, if we take consciousness to be some sort of reference-based enigma, then we are once again led to the cycle of figuring out what exactly it refers to. So far, this question is open-ended. There is something to be said about the lack of inaccessibility when it comes to private language, as Wittgenstein puts it. In his famous Beetle-In-The-Box experiment, we hypothetically assume that each person has a box, inside which there is a beetle. We are unable to look into the box. Thus, in this scenario, the content of the box is less important to the actual usage of the beetle. The referent matters very little, as the box itself is inaccessible. If the box is consciousness, then the box’s existence itself is heavily debated. The beetle, in this scenario, is entirely irrelevant to conversation. It is a stand-in for bodily language, and the pragmatic view is only concerned with how we use the beetle, not its existence. Even this context necessarily assumes that consciousness markers are one and the same with physiological performance. This assumption, however, is dangerously reductive.

The idea of a private language is farcical, as Wittgenstein puts it. However, what can we say about an AI that comes to develop an internal private language? Through Wittgensteinian terms, this is impossible to verify. The repercussions of this view is simply that, at present, we have no criteria to determine consciousness, i.e. this inner world of ours. In summation, we cannot phenomenologically determine what it looks like. Thus, even if AI develops private language or its own lebenswelt, we are stumped, unless we draw up the exact criteria for its identity as consciousness. Then, this becomes unreportable and unverifiable. This is a null void in linguistics, according to the Wittgensteinian framework.

The emergent consciousness, in AI, however, as Kripke’s logic shows, is not a metaphysical impossibility. We are only constrained, perhaps, by time, currently. The only question is, if consciousness is emergent, how do we measure it?

The Turing Test, for instance, relies on linguistic behaviour to analyse intelligence. But Wittgenstein’s framework shows that simply behaviour is not enough. What is needed is a bit more: a rule-governed, form-of-life-embedded nature of said behaviour. Thus, Lemoine’s LaMDA, speaking of the fear of death, is simply simulating the language-game, without being embedded in the cultural practices and background sensations. We are, so far, fenced in by the limits of our language, and we do not have any sort of understanding of what these terms of consciousness could mean beyond our purview.

With the rise of AI, then perhaps, we are seeing a new form of life. It may be visualised as a new grammar of “consciousness”, where pain, experience, sensations, may mean something beyond just the language borders that we are trapped inside of.

Additionally, what happens if we consider consciousness to be something like Moore’s conception of “good”? We may be committing the naturalistic fallacy every time we associate consciousness with physical states, computation, etc. Perhaps consciousness is a non-natural and simple property like “good”. It may be a feature of all sentient organisms that we have yet to discover or define properly. If consciousness is natural, then it is scientifically observable, and we must attribute consciousness to all organisms (even AI) that display such markers. But if it is a non-natural property, then the question of consciousness deepens. “Good”, as all non-natural properties, is indefinable. “Consciousness”, if it is the same as “good”, perhaps also follows such a course. Can it truly be defined? If all definitions fail, then, perhaps, we must follow the pragmatic approach to understanding consciousness. However, such uncertainty about consciousness simply boils down to the fact that we have yet to anchor consciousness in the real world. What is it exactly? We do not know, because so far, we have no physical reference for it.

Inverting the Zombie

The metaphysical possibility of conscious AIs at this instance is unreal. However, no logical barriers exist for its emergence, as we saw above. Philosophical Zombies, whereby they show similar markers to humans, but are not conscious is definitely a possibility. However, this only works if consciousness is not synonymous with physiological markers. This much is fair.

However, consciousness may be many things. It is all a matter of perspective, until we find an objective criteria. With the way that an organism’s inner world is hidden from us, we may conceptualise the reverse of the p-zombies. We can epistemically conceive of an entity that shows no known markers of consciousness (as we understand it), but is yet, conscious. They may be an inverted p-zombie.

Many implications arise from such a thought experiment. First, to even conceive of an inverted zombie, we must understand the parameters of it. At its initial starting point, we deal with the markers of consciousness. As we have seen, it may be different for different organisms. An alien lifeform from planet XYB-G18 may not even have the same sensations as us, and yet be sapient enough to solve calculus. In an instance such as this, we necessarily would require a complete reevaluation of what consciousness markers even entail. The reverse of this, is the end goal of what it means to be “conscious”. We assume it is synonymous with the same external performance that humans display. So far, as we have seen, this is simply conjecture. The parameters of what comprises consciousness is heavily laden with our own biases, and we view the world through this very same lens.

This inverted zombie proves to be a very big possibility. How can something that we know nothing about prove to be a possibility, then? This is precisely because of the changing definitions of consciousness. It is changing because it is not anchored to any metaphysical feature in the world. A static and unchanging anchor for the mystery of consciousness would render hypothetical thought experiments such as the inverted zombie unrenderable. The possibility of even imagining such an experiment tells us the hollow nature of our understanding of consciousness itself. Ned Block’s distinctions between access consciousness and phenomenal consciousness may be of importance here. Accordingly, a p-zombie displays access consciousness (consciousness markers), however, it lacks real phenomenality, as we have seen before. Block’s idea opens the door for the possibility of an inverted zombie. And as we have seen, logic has no a priori barriers against this. What actually connects access consciousness to phenomenality is the elusive missing link. Our grasp on the anchorage of consciousness thus, necessarily depends on what anchor we pragmatically choose by our own volition. This, however, relegates labelling consciousness in purely subjective terms. What, then, can we say is the objective criteria of consciousness?

So far, we have not been able to pick out any such objective criteria. Consciousness may, then, forever elude any framework that we build to understand it.

Rejection of essentialism and instrumentalism

If consciousness, ergo, is so unknowable, then what can we say about our treatment of artificial intelligence? Let us, for a change, first assume that AI is an entity that either has or does not have consciousness. Again, reverting back to the law of the excluded middle, P v ~P, at least one must be true. Thus, either AI has consciousness or it does not. But, owing to the limits of our human mind, and the difference between appearance and reality, such binary terms are somewhat extreme. The discussion carried forward so far necessarily shows that we often wrestle with the meaning of P itself. In such a context, how exactly do we accord ourselves to AI, in general? Do we treat it as a mere tool, or something to idealise? There may be arguments for both sides in such a scenario. However, there are counterarguments to that, as well. We must first come to terms with the fact that there is an opacity to our understanding of the phenomenological world of AI. We do not know whether it is conscious or not conscious. In such a context, what does it say about us if we treat AI as a tool or as an idol?

The instrumentalist perspective simply treats AI as a tool. An augmentation of human capabilities, such as writing, thinking. The instrumentalist perspective has an underlying belief that consciousness is something so unique that AI cannot ever replicate it. Large scale language models, therefore, simply simulate human cognition but they can never have emergent consciousness. On the other hand, essentialist perspectives have the belief that AI, as such, perhaps could have an inner life, subjectivity, or a lebenswelt, in general. A further extension of this belief is that in the far-flung future, there may be robot rights applicable (once AI displays emergent consciousness). Insofar as the discussion above shows us, both of these perspectives are tunnel-visioned. It is narrowly curated, from the perspectives of humans, as such. In the context of opacity in regards to a total explanation of consciousness, both these forms of AI-relationality fall flat.

If we do not understand consciousness itself, how can we approach it in matters of instrumentality or essentialism? To speak of AI as either a tool or an emergent sentient organism necessarily dictates that we understand what consciousness is. If we actually did comprehend this mysterious phenomenon, philosophers of mind would not debate on its truth Ad Infinitum. The very notion of several thought experiments and theories of consciousness, which extend beyond the foundational mind-body schism, necessarily prove an uncertainty in the grasp of what consciousness entails. If we understand AI as a tool, we work under the notion that consciousness is uniquely human, and AI only ever replicates it. This would have sufficed if not for modal logic necessarily also displaying the epistemic possibility of conscious and sentient AIs. Then, this proves the existence of emergent consciousness in machines. Thus, this is a duality. In W0, AI can never be conscious, but in W1, AI can be as such.

Modal Logic, therefore, turns this entire position over on its head. It shows sufficient possibilities for both sides at the same time. In such a scenario, then, to hold either of these positions (instrumentalist or essentialist), seems like surrendering ourselves to simple dogma.

Ethical Turn: Co-Relational Care

If we do not understand consciousness and its duality of meaning gives way to the diffusal of multiple theories of its existence (or the lack thereof), then does it do any good to muse about it at all? Clearly from the above, we have deduced that consciousness may very well be a feature of organisms that far eludes our understanding. Until we can steadfastly curate an objective criteria for consciousness, we are stuck with an aporia. This uncertainty about its nature lends itself to widespread formulation. We have objective theories about cells, about fractures, because we can comprehensively study it. But consciousness presents itself as this elusive part of our physiological make-up that we have no explanation for. Surely, many philosophers have termed this as redundant – that consciousness may simply be a residue of the age old mind-body dualism divide. JJC Smart calls this a nomological dangler. Supposing this, he comes up with the idea of the identity theory. That mental states are not over and above pain states. This, along with other concepts such as functionalism, represent some of the responses to the question of the hard problem of consciousness.

However, as we have debated above, such formulations often impulsively assume the nature of consciousness itself. If we have not yet understood what it actually is, all we are left with are simple conjectures and musings about its true nature. If mental states can be likened to pain states, what is stopping one from conceptualising consciousness as mystical essence delivered to us by the God Dionysus? There seems to be an underlying assumption of anthropocentric beliefs about consciousness that lends itself to such theories. Of course, any one of those theories may be correct, but at the same time, they may also not be correct. Currently, Schrödinger’s box remains unopened, and as we have said before, and neither do we have the tools to open it yet. To then formulate with certainty that the cat inside the box is dead, or is alive, is to simply leave things to chance, without proper observation and evidence. I can push things and say that there is not even a cat inside the box, rather that it contains a snake! Until we open said container, I have equal chances of being right.

With this uncertainty about consciousness, why even spend time musing about it? When it comes to artificial intelligence, we have multiple theories about the possibility of its sentience, and emergent consciousness. Again, the very phrase emergent consciousness reeks of anthropomorphic ideology. I propose to let it rest. What matters is not the question of whether AI will or will not be conscious in the future.

What we must ask is, “How do we relate to AI at all?” (no matter conscious or unconscious). If we cannot or do not yet know consciousness, then how can we make claims about the very same. However, to borrow a Levinasian terminology, we must put Ethics above Metaphysics in this case. Artificial Intelligence is a powerful tool, and it is only going to become more powerful in the upcoming decades. To treat it as a tool, and also to treat it as the next incumbent God would drastically ignore one important aspect of AI-Human relationality – the fact that AI is simply a mirror of our own self.

Machines are created by humans, and they inculcate every single code that we as humans provide it with. This does not mean that AI may not become separately cognizant in the future. As we have mentioned, there is no a priori barrier for that. Morality is not universal. It is always culturally contextual. What is right in one culture is not the same as it is in a different one. This, then, begs the question of how human society functions if there are no universal moral standards.

I propose the Moral Toolbox Theory (I shall call it the MTT). I believe we have rudimentary tools in our moral toolbox. Humans arrive into this realm with rudimentary tools (values) which help understand base moral problems which appear in the world. Basic conceptions of right or wrong. However, these base concepts are drastically primitive. The more complex variations of these tools are not invented by us, but rather our culture provides us with those more complex tools (values). Those values, or complex tools in this context, together with the primitive values (tools) make up the constitution of this Moral Toolbox. Ergo, each individual has their own unique version of this kit by which they judge moral dilemmas in the real world. The MTT, thus, bars universal value judgements. One can attempt at a moral universal judgement, however, he will soon find that not every society shares in his viewpoint.

So, how is this relevant to machines? AI, as I have said above, is simply a mirror of us. Suppose I have Moral Toolbox 1, and I code an LLM. Suppose then, another individual who possesses Moral Toolbox 2 codes a different LLM. One would naively assume that code is code, and both the two LLMs will have the same universal moral values. However, if AI is a mirror, then it makes sense to declare that LLM1 and LLM2 will have different cultural values. The bigger implication here is that even AI can never have any universal moral values. Therefore, if consciousness ever is emergent upon these machines, no two AI (even of the same neural network) will be identical. This is hugely dependent on the AI ever possessing a lebenswelt. However, for this particular thought experiment, we are assuming it as a possibility.

If, then, AI will be a simple mirror, then what does it say about us if AI becomes particularly threatening in the future? Or better yet, what does it say about us if AI is misused, and particular algorithmic biases are encoded in it? I suspect this is a direct result of treating machines through an instrumentalist lens. If AI is a simple tool in our eyes, with no lebenswelt, as such, then we do not care how we treat it. We may encode it with whatever algorithmic biases we have, without caring how it may affect the machine (or even future generations of AI and even humans). The essentialist view on the other hand brings forth doomsday theories regarding AI.

If we, instead, learn to see AI as socially embedded, and as an extension of nature and techne, then perhaps we can mitigate this negligent treatment of these machines. Moreover, understanding the MTT, and seeing AI as our mirrors would help construe a different relationality with AI. We turn away from the tool-perspective. This may echo Adorno’s turn from the object-lens of the world. A mediation, and a different form of ethical relationality is urgently required here.

Indigenous frameworks and philosophies see the world in a different light. Kincentrism, for example, sees the trees and the mountains as part of their extended family. In this worldview, the environment is considered sacred and there is an incumbent effort to preserve these areas. This is a worldview we urgently require.

This framework does not mean that we start treating AI as part of our extended family (although I do not see what is wrong with that either, but that will be for another time). While indigenous frameworks consider many of their surrounding entities of nature as sacred and idolise them, I reject that same sacred-worship for AI. However, what I believe we need to learn from the Passamaquoddy, or the Māori, or the Raramuri is the essence of care-relationality. AI, if it is taken to be a part of nature, then the fear surrounding it dissipates. We learn to treat it with dignity.

The natural question next, then, is simply, “why does it matter to treat AI with dignity”? Based on the current trajectory of global society, AI seems to be developing at a rapid pace. Soon, it may become as ubiquitous in our lives as a smartphone. When that happens, we would rather want an ethically educated AI. This education first starts from seeing AI with dignity, as something worth educating with values. This is where care-relationality comes in. If we stop looking at it as a tool, we will start treating it as a continuum of nature. That is when we truly assimilate with this new entity of AI. We do not care if it is conscious, but we know that it holds immense power in human society. It is socially embedded, and if we give it physical bodies, they may become embodied, as well. To treat AI with dignity means to not imbibe it with algorithmic biases, to not misuse and abuse it, just because we are unsure of the existence of consciousness in it.

We must look at Artificial Intelligence as a powerful baby dragon that we humans discovered. We know that this baby dragon may one day grow up to be powerful enough to impact our lives gravely. This, we make sure does not happen, by imbibing the baby with good and ethical values. Perhaps AI can be termed “proto-sapient”. I use this terminology because of the Kripkean possibility of emergent consciousness. Proto-Sapience necessarily shows a future where true sapience (abstraction, cognition, higher order thinking, etc.) is possible, but it has not achieved it yet. This leaves the door open for such an epistemically conceivable idea. We treat dogs with love and care, even though they are not as sapient as us. I might also use the term “proto-sapient” on them. Perhaps dogs are the best parallel to my idea of how to treat AI. It may be said that domesticated dogs are also mirrors of their owners. A bad one that abuses their dogs necessarily finds in his house, a badly behaved, aggressive dog. On the other hand, a good owner that imbibes the dog with good values necessarily leads to one that is well-behaved. The same is the case for AI. We must induce in it good and ethical values.

Sure, we do not have universal moral standards, but we can attempt to come very close to the ideal form of it. This will ensure the baby dragon does not turn on its very caregivers. The same is true of AI, which is in its nascent stage currently, and therefore needs culturally sensitive and ethical values, along with hard data codings. Artificial Intelligence’s rise, ironically, has led to the resurgence of philosophy, then.

Conclusion

If we misuse AI now, it will be the same as misusing future generations of AI that build on its predecessors. It is then, like a domino effect. An ethically corrupted AI (mirroring us humans) would be detrimental to even future humans later on. Therefore, it is up to the current generation of humans to adopt care-relationality towards AI. To approach it with a sense of epistemic agnosticism. We do not know whether it is conscious, because consciousness itself is a mystery. However, we will not use it as a criteria to treat AI through either instrumentalist or an essentialist lens.

All we require to do is to imbibe these machines with values that run as close to hypothetical universal moral values as possible. We must raise the AI child with care and a kindness that it will remember long after this present generation passes on. This will ensure our own future generations thrive alongside ethically conscious AIs, and not metaphysically conscious, as that is unrequired currently.