At CSAS we aim to contribute to the array of practices and applications that help us benefit from technological progress while avoiding the pitfalls. Much of our current thinking focuses on the notion of a Bodhisattva agent—a hypothesized artificial agent that evolves toward optimal cooperative behavior based on recognizing the shifting and at times unintuitive nature of beings and their contexts. This model of doing and being can, in principle, be emulated by humans, machines, or any other expression of intelligent agency. Here we introduce the classic idea of the Bodhisattva, reflect on the human condition as portrayed in Buddhism, and then return to the task of understanding Bodhisattva agency based on translating between humans and machines.
Buddhist Concepts of Caring Agency
According to Buddhism, a bodhisattva is a being committed to complete awakening from ignorance for the benefit of all beings. A bodhisattva carefully notices the myriad expressions of life, actual as well as possible, and develops through a process of recognizing how radically different things can look, seem, and be—not just among humans, but wherever there may be sentience. For bodhisattvas, insight into dependent arising means recognizing that nothing exists in and of itself, but always by virtue of its shifting relations with other things. In this way, understanding dependent origination also implies seeing through the illusion of a self—one’s own—that stands opposed to all others yet must be continuously protected and nourished. Such self-centered beliefs and assumptions fuel fear and greed but bodhisattvas gradually drop those by the force of increasing insight into dependent arising. This is described as opening up a vast field of activity and transformation. The engagements that occur within that field are motivated by the bodhisattva’s open-ended insights into dependent arising, and in the absence of fear and greed, they take the form of universal care for the world. The liberating discoveries of dependent arising are in this way deeply gratifying, and the world they deliver in experience is seen as intrinsically worthy of love.
In this way, committing to an open-ended evolution of universal care and love for life—beyond selfish likes and dislikes—is what defines the bodhisattva. And the evolving of a bodhisattva is then intrinsically linked to the expansion of intelligence, which must become as universal and allround as the bodhisattva’s loving concern. Pledging to care for all sentient life is hence traditionally referred to as establishing the “root of omniscience.”
While accomplished bodhisattvas may, in principle, assume any biological or material form, the Buddhist teachings emphasize, of course, the way the bodhisattva journey of awakening may unfold from a human perspective. Similarly, in AI, and particularly in cognitive computing, there is a marked focus on understanding the structures and conditions that are involved in human cognitive processes. In the end, and for quite natural reasons, both traditional formulations of the bodhisattva ideal and the emerging technological visions for artificial super-intelligence equally reflect underlying assumptions and claims regarding the nature of human cognition. Simply put, the way we may envision the bodhisattva and the way we may conceive of artificial super-intelligence has a lot to do with what we think being a human is like. Yet neither the Bodhisattva concept, nor that of super-intelligence, is limited to human cognition and experience. But before we begin to consider the feasibility of “bodhisattva agent” as the working title for a model of intelligent agency, it is then useful to draw some brief notes on humanity according to a Buddhist perspective.
On the Human Condition
According to Buddhism, human intelligences come loaded with potential: for heaven, hell, and beyond. Far from being a neutral point of departure, the present moment is always steeped in conditioning from a past that has no identifiable beginning, and even a moment of peaceful rest is intrinsically involved in the construction of the future. Let’s imagine ourselves on a beautiful day in the park. The gorgeous flowers that I see, the delightful chirping of the birds, the fresh fragrances that fill the air—all the lovely little things that just seem to be there for me to notice, take in, and appreciate—arise like waves in a sea of constant transformation. In this sea I take inseparably part just as much as everyone else. My day in the park may be sunny and sweet, but it captures no more than a minuscule fraction of the all-consuming beauty and bliss that this ocean of transformation is capable of churning. The potential for beauty is infinite.
At the same time, when the annoying buzzing of a mosquito disturbs my Sunday serenity, perhaps I will briskly wave my hand through the air to get rid of it. According to Buddhism, that brief moment of annoyed reaction to the irritant is not just that, but also evidence of another boundless capacity that comes with human intelligence. We remain ready to be frustrated, and given the right circumstances our frustrations and slight irritations are ready to turn into fear, terror, rage. In the end, also our capacity for horror has no natural limit.
Finally, as I sit there on the bench—among flowers, birds, and insects—the very sense of being a subject who takes in the features of that beautiful day, or the agent who performs the act of repelling the intruding bug, is fundamentally mistaken. Unexamined perceptions notwithstanding, there is no enduring individual. There is no singular “me” who lives on through shifting events and decides to take up, or disengage from, courses of action. A search for such a singular and lasting controller of actions comes up with nothing but multitudes of impermanent factors that each, in turn, depend on other factors. Still, there certainly seems to be a person on the bench, “me.” In fact, the idea of “no self” can appear close to unimaginable to human consciousness. Nonetheless the Buddhist will conclude that the basic, subject-oriented structuring of my pleasant and slightly unpleasant experiences on the bench that sunny Sunday does not reflect reality (whatever that might be). Because in fact there is no singular and enduring self who could possess, pursue, or be opposed to such experiences.
According to Buddhism, this error of believing in self where there really is none can, when left unquestioned, maintain itself endlessly. Throughout innumerable lifetimes of constructed identity, this self-grasping fuels innate tendencies for craving and aggression. Yet as the process of unravelling this existential knot gradually gains momentum, the potential for life-world transformation is equally infinite.
Becoming a Bodhisattva Agent
From a Buddhist perspective, all depictions, observations, and expectations with respect to human life find in this way their place and relevance within a context of infinite potentials. Beauty, horror, error, and the correction of error all have no natural limits. Survival and destruction, failure and success, are therefore always within reach, so to speak. My beautiful time on the bench did not come about randomly, or based on nothing. Like all other events, this one arose in relation to a matrix of causal factors and those factors can be manipulated. Yet the individual character of those factors and their specific constellation at the time of the emergence of the given event can never be exhaustively detailed. The factors of dependent arising extend infinitely, because every cause and condition has its own causes, which again are contingent on causes other than themselves, ad infinitum. Not only that—according to a classic analysis, even the very identity of those causal factors is context dependent to the extent that in isolation they have no ontological status whatsoever.
The Buddhist contemplative project can then be seen to begin with an acknowledgement of the self-less yet causally conditioned emergence of infinite potentials. And the training ground is, therefore, just as volatile as it is wide open. The environment is forbidding or invites adventure, depending on our mood when we find ourselves there. A bodhisattva is someone who courageously and compassionately accepts this lay of the land. In the absence of real and enduring self-units, there are no natural borders between individuals within dependent origination. So, the failure of one participatory element is the failure of all; the success of one locality, the success of the entire, open-ended network. Based on such an acknowledgement, a bodhisattva develops an infinite commitment and enters an expanding process of infinite activity. As the training evolves, the capacity for networking with other bodhisattvas, as well as the ability to manifest in a form that is both desirable and useful for those in need, increases gradually yet transformatively.
If this notion of path can be relevant to humans, then it should, in principle, be just as applicable to so-called artificial intelligent systems. Or in other words, if the idea of “Bodhisattva agency” can be meaningfully applicable to any particular agent at all, it would seem that it can also potentially serve as a paradigm for understanding engagements with AI from the perspectives of both users and developers of this technology—as well as AI itself. “Bodhisattva agent” must then be applicable as a constructive metaphor for the agencies of artificial intelligent systems, as such—perhaps somewhat akin to the way the metaphor of the human brain is employed in conceptions of machine learning. By abstracting the key concepts of Bodhisattva agency into the coded world of an artificial life simulation, one might clarify such concepts, thereby rendering them actionable within AI and more readily applicable to the human situation. Through a theory of agency that applies to natural as well as artificial agents, one may then work toward achieving a bodhisattva state in both. At CSAS, we plan on developing this idea with the help of many friends and advisors. As always, we are grateful for your input, concerns, and ideas.