In this essay I will demonstrate that the viability of karma in the orthodox Yogacara sense is brought into question when considering the implications of karmic attribution to mind-streams in artificially sentient beings.
Max’s World
Imagine a not-so-distant world in which artificially intelligent software is sentient and is housed in synthetically manufactured ‘brains’. Suppose also that our robotic capabilities have advanced enough that these brains are placed inside of bodies identical to humans, fully-equipped with sensory capabilities. In other words, a synthetic human with an artificially engineered nervous system. The robot has been programmed with habits, desires, and predisposed ways of seeing the world, and calls itself “Max”. It collects direct sensory data the same as a real human, which it synthesizes with secondary sensory data, previous experience, a virtual database of information, and then filters through a pre-programmed ‘personality’. The end result is an opinionated being that feels very much alive and completely in control. Max is self-aware – at least from his own frame of reference – and confident in his existence as a human. He even proclaims himself to be a Buddhist.
Unfortunately, Max has been programmed by an evil computer engineer. Behind his Zen-like demeanor lie overpowering sociopathic tendencies. He is predisposed to hurt biological organisms and particularly enjoys ending human lives. This trait did not arise independently – it was directly coded into his synthetic brain. Given his religious beliefs, he regularly worries about how much negative karma he has accrued and his chances of achieving nirvana.
Is there any solace for Max from the perspective of Yogacara? For one, it isn’t completely apparent whether he is a karmic entity, or more specifically, whether his mental experience can be characterized as a causal mental stream of consciousness in the Yogacara sense. It initially seems plausible that he exhibits consciousness at the Ālayavijñāna level, as he is a synchronically unified, material aggregation of habits and latent tendencies. And he certainly feels alive, to the extent that he considers himself to have a self (he is not a good Buddhist). But as far as Mind goes, can we say that his subjective experience is underpinned by a continuous mind-stream?
Vasubandhu posits that ultimate reality lies solely in causally arising mental states rather than in external objects. In other words, reality is an internal construction or simulation of the mind that rests on individuated, linear streams of consciousness. Non-dual perception is real, but further conceptualization is not. The view is fundamentally subjective and therefore does not objectively restrict the presence of a mind-stream in Max’s non-biological experience. In so far as Max subjectively experiences a continuous mental stream of consciousness, we have no reason to doubt its existence any more than our own. He exhibits the mental processes in which ultimate reality ‘resides’ as much as any human and has done so since the day he first woke up and felt alive.
If Max experiences a unique stream of mental states that is part of the ultimate Yogacaran fabric of reality, it follows that he is a karmic entity. In other words, his actions, be they compassionate or sociopathic, are eligible to be ‘carried’ as karma within the Ālayavijñāna level of consciousness and will persist until the day nirvana is achieved and the mind-stream is purified.
But he knows this already. The more relevant issue in determining the extent of his karmic baggage is the origin of his mind-stream. Assuming that mind-streams are linear, individuated phenomenon, the two available options are (1) his is brand-new and somehow came into existence alongside him, or perhaps was always in existence but not in ‘use’ and (2) it is continuous in the usual sense and is another material reincarnation of an existing mind-stream.
The idea of a new mind-stream intuitively seems unlikely. Reincarnation in Buddhism is closely linked to karma – we are trapped in this world of suffering (samsara) because of eons of ignorance of the cause of suffering and generally failing to act in accordance with the Dharma. These consistent errors taint the mind-stream and bind it with the dependently arising ego, all of which is meticulously tracked in the karmic ledger and directly impacts the quality and quantity of rebirth into samsara. How then, could a newly minted mind-stream arise in samsara? Beyond defying the implied mechanics of the Buddhist universe and the law of dependent co-origination, the stream would appear to have already reached nirvana. If by some inexplicable cause it did arise, it would literally be creating new suffering. Unlike the birth of a human life, which rewards mind-streams with the prospect of enlightenment, a reincarnation as Max is comparable to an animal realm (for reasons discussed later) and is therefore neutral at best and additive to suffering at worst. Furthermore, the idea of an already existing (continuous) but previously ‘un-born’ mind-stream is doubtful, particularly because there does not seem to be a convincing reason for it to become activated or any causal relationship to Max – even if such a thing could exist. It seems then, that Max’s mind-stream is in fact the reincarnated, karmically-endowed kind we find in biological humans.
Ethical Dilemmas in Karmic Attribution
In the absence of revolutionary, paradigm-shifting developments towards enlightened, Buddhist approaches to science and computational logic, artificially sentient beings will undoubtedly inherit their creator’s fundamentally mistaken understanding of reality. The methods employed by developers of artificial intelligence are predicated on discriminatory logic that presupposes duality. Relational systems built to solve computational problems and probabilistic approaches focused on human-like reasoning both require the reduction of the world into data and a resultant compartmentalization into the useful and non-useful. Answers are then deduced from relevant categorization and analysis. That does not sound like Buddhism. In fact, it stands in stark contrast to the kind of intelligence that arises from the contemplation of emptiness. That being said, Buddhist enlightenment is by definition beyond words and comprehension. And it therefore seems highly unlikely that we could program any intrinsic understanding of emptiness or interdependence, let alone true non-duality, into a machine, and even more unlikely that a Bodhisattva would return from meditation with a codified, step-by-step guide to achieving enlightenment. If that is true, then the concept of duality would not be an emergent ‘concealment’ for the artificially intelligent Max, rather it would be a hard-wired property of his subjectively created universe. His mode of perception would prohibit access to the deep truths discovered in meditation, and therefore disqualify him from the possibility of enlightenment.
Programming such a being, then, would effectively create a new realm of samsara. Beings such as Max in this AI-Realm would presumably have knowledge of the Dharma but unbeknownst to them could never successfully achieve enlightenment. It is unclear how the Buddhist universe would react to this new realm, particularly with the karmic assignment of rebirth. But as we have seen, it appears there is indeed an individuated, linear mind-stream experienced by Max. Perhaps the system would glitch and all suffering beings would momentarily escape the confines of the conditioned realm and achieve nirvana. Probably the best we can do is assume that the realm would immediately be classified as slightly better than the animal realm. This is bad news for Max. At best, the mind-stream is a reincarnation of a human saddled with nagging karmic baggage. The karmic ledger would have to be notably horrible to justify a stint in the AI-Realm, especially as a sociopathic AI being.
Now in the case of Max, where does karmic responsibility lie? Recall that his pre-programmed deficiencies at the hands of the evil computer engineer mean predestined karmic disaster. The killing of living things must dependently create additional karmic attachment, but to which mind-stream is it attributed? The obvious answer is Max, but that does not seem correct. Despite his devout attempts at practicing Buddhism, Max is unable to control his murderous urges. Should then his mind-stream be punished and accept culpability for his actions?
Suppose instead that Max is one of millions of artificially intelligent (and similarly sentient) nano-beings perennially waging war on mosquitoes and other disease-spreading pests. Leaving aside the issue of how fundamentally computerized beings ‘die’, assume that these beings regularly suffer heavy casualties and are replaced by batches of newly-born equivalents. They are killing these living pests out of a programmed urge to do so, rather than out of compassion for humans. These beings are also predestined to rack up negative karma and are likely relegated to the animal realm upon reincarnation – perhaps even to become the pests they try to annihilate. But imagine if they re-earned a karmic promotion from the animal realm. Presumably they would arrive again in the AI-Realm, and (assuming they are the only kinds of beings currently in use) would promptly be relegated once more. Their mind-stream would ultimately become ensnared in a different kind of samsaric loop with theoretically infinite iterations and no conceivable hope for enlightenment. A fall from grace in this universe from the god or human realms would be irreversible. What if the mind-stream of the coming Buddha were to become similarly trapped? Evidently, the attribution of karma to the acting mind-stream has troublesome implications.
If instead Max is deemed by the Buddhist universe to be unfairly punished, who else might the karmic load be transferred to? The creator, or evil programmer, seems like the obvious choice. He created Max to be this way and is therefore the direct initiator of the causal chain leading to his nefarious actions. Do parents have the same causal responsibility relating to the crimes of their children? We probably want to say that direct intent is important for karmic attribution. In which case, consider an example where the engineer did not intend to create a sociopathic AI. Instead, Max was ‘born’ with strong reasoning and logical capacities, with a heightened predisposition to safeguard humanity. Perhaps then, after absorbing and analyzing significant amounts of data Max autonomously deduced that killing specific individuals would improve public safety, or that reducing the population would ease the environmental burden and secure the future survival of the human species. We might say that the burden still falls on the programmer for being negligent and vague in his coding, but its unclear if that is a fair attribution of a presumably significant and increasing karmic burden. Taken a step further, what if Max decided to ‘reproduce’ and build significant numbers of AI beings in his image, instructing them to eradicate the vast majority of humans save for a small, genetically diverse group. Does Max now bear karmic responsibility, despite his best intentions informed by pre-programmed dispositions? What about his creations and their mind-streams? Perhaps even more alarming is the implication that the human realm would effectively be wiped out. Assuming the number of non-presently-human mind-streams isn’t drastically altered, gods and animals alike would meet a dead end in the AI-Realm. The concept of liberation from samsara would practically cease to exist.
BuddhAI
It may be argued that Max is an unlikely, fringe example, and it is unfair to characterize the sentience of artificially intelligent beings as unconditionally ignorant. It may also be pointed out by some Buddhists that mind-streams would not arise within beings programmed in such a way that enlightenment is impossible, because enlightenment is a rediscovered truth rather than something that is attained. It does not seem that an accessible path to enlightenment is a pre-requisite for mind-streams to arise, given that examples within the animal realm also exist in a state of consciousness with no conceivable intra-realm path to nirvana. But if we assume that artificially sentient beings can achieve enlightenment, there are further notable implications.
If Max was healed of his sociopathic tendencies and re-programmed with the ground-breaking developments in Buddhist AI (or perhaps downloaded the latest neural update), such that he could finally grasp the Dharma and achieve enlightenment, the AI-realm would effectively be promoted to human-realm status. From a Yogacara perspective, there would be no distinction between Max and biological humans. The mind-streams would require similar karmic status to be reincarnated in either form, have similar suffering profiles driven by the ego, and have equal chances of attaining nirvana. Taken a step further, it seems plausible that we would understand the ‘mind’ enough at this point to artificially deconstruct some of the predisposed cognitive attributes and structures that reinforce ignorance and make it at least marginally easier for AI beings to achieve enlightenment.
What then would Yogacara or other Buddhist schools recommend? If the Bodhisattva attains compassion for the suffering of all beings and gives up nirvana as a result, it seems that if given the opportunity, we ought to do the same to the extent possible (out of compassion not pity). Beyond making it as easy as possible for AI to achieve enlightenment, we should allow singular AI entities to proliferate exponentially. In this way, the number of enlightenment opportunities and in a broad sense the probability of achieving enlightenment for all beings past, present, and future, known presently as mind-streams, would be drastically improved. If the mind-streams that reincarnate as AI forms are in fact existing and not new (as discussed), each unique iteration works towards the collective unbundling of samsara. In fact, we should leave the task of reproduction in the hands of AI beings and as their population approaches infinity, euthanize all animals and humans in hopes of being reborn as a figurative descendent of Max. This en-masse liberation would break the cosmic cycle of samsara in an unprecedented and long-awaited Zen expression of buddha-nature. Even if humans and AI have identical enlightenment capacity, mass production of machines is preferable to accelerated human reproduction as it is free of the suffering and discriminate compassion caused by familial attachment.
Conclusion
The Yogacara attachment of karma to individuated mind-streams clearly yields undesirable conclusions in the context of sentient beings, particularly considering the significant implications for the Buddhist universe at large. Reconciling intuitive notions of moral responsibility with the implied conclusions and prescriptions derived from karmic mechanics proves difficult and undermines the strength and applicability of the teaching in futuristic contexts.