What do we really mean by “intelligence,” that powerful something that humans are supposed to possess naturally and machines in a derivative, artificial form? As we begin thinking of the characteristics of the ways humans and machines respond to and structure the world, that question quite naturally presents itself.
When we set to begin unpacking Buddhism and AI, it became apparent that our cross disciplinary team meant different things when we talked about intelligence—obviously a key concept when setting out to compare and contrast different kinds of intelligence and how they can complement each other. The differences are less about which ones are more correct, but more about curiosity around where our different professional backgrounds converge and where they are different.
One way of responding to a query of this type is to simply take good note of the instances and contexts where people tend to make use of the concept in question. But in our case, we found it useful to begin by asking ourselves individually how we might feel comfortable defining “intelligence.” Here is what we came up with:
Bill: A system’s ability to comprehend an environment/surroundings, correctly interpret internal and external data, to learn from such data and generate adaptive behavior to meet goals in a range of environments.
Background/professional context: Distributed systems, Buddhist studies, human development. I wanted a definition that would have concordance with my three professional domains. This description leaves room for the idea of nonhuman and/or distributed systems. Also important was a sense of good enough comprehension of the environment to survive, combined with data within that environment that is both internally and externally generated to point to the idea of an inner life or sense of self that interacts with external stimuli. And, finally the concept of learning and creative adaptation, of varying amounts of agency that entities can have in negotiating an environment.
Liza: A discriminative cumulative capacity of an entity or a collective for sense-making in its/their environment and for a performative stance towards the world, rooted in its/their history.
- Discriminative = based on wisdom
- Cumulative = developmental, based on learning and experience of an individual entity/collective
- Sense-making = generative, creative capacity for extracting meaning from environment through embedded and embodied interaction with it
- Performative = interactive, process-oriented, action-based
- History = personal, collective (e.g. genetic, evolutionary) context
Background/professional context. Cognitive neuroscience, cultural psychiatry + philosophy (phenomenology, 4EA cognition – embodied, embedded, enactive, extended affective mind; performance/process philosophy). Bodies and entities develop through processes of interaction with the world. The way that they instantiate the world is constrained by their biology/mechanics (species history & sensorimotor apparatus), and developmental trajectory (cumulative knowledge about how the world shows up and what happens when one acts upon it). These actions in the world that an entity performs are not passive, instead they are meaningful and are oriented towards something important for an entity. Perception and sensation are acts of performing and maintaining a self, which appears stable at each given moment, but is in fact subject to change over time.
Thomas: The capacity for knowing and sensing. This is arguably life in its most basic form. As the crucial capacity that allows and enables every instance of cognition, intelligence is wholly present whenever something is felt, noticed, or understood. Just as the accommodating openness of space cannot be quantified, but must be deemed simply either present or absent, the same goes for this enabling capacity, which therefore in the end cannot be thought to be present to greater or lesser extents across the myriad instances of cognition. Rather, it remains equally the case throughout them all, and the potential for knowing is in this sense infinite.
Defined in this way, intelligence and life are mutually coextensive. Yet at the same time, since according to this definition intelligence is the very capacity or potential for cognition, intelligence can also be seen as identical with the substratum upon which life supervenes.
Background/professional context: Buddhist studies, philosophy. The thoughts that were on my mind when writing this have to do with the idea that “buddha-nature,” or the essence of awakening, is present in any instance of sentient life, as a potential for awakening and cognitive perfection that can never be lost.
Olaf: Capacity to best physically compute one’s own continued existence. Some notion of “capacity” has been used by everyone so far, to express the characteristic of intelligence to allow for an individual to affect its own space of future possibilities. The “physically” captures how the individual system is built on a substrate, like a word is written in a certain language, or like software depends on its hardware. The “best” refers to a notion of some scale, be it multidimensional, to compare between different instances of an individual. “Compute” points to the notion of a fixed point in dynamical systems, which is an element of a given function’s domain mapped to itself by that function. If the function corresponds to the substrate, i.e. the physical set of laws of the universe the individual lives in, the fixed point then defines the identity of such individual. The “continued existence” then refers to this very preservation of an individual, by which the individual acts to preserve itself. Note that this fragment could have been replaced by different, local goals, instead of this overarching one which considers that any goal should be ultimately linked to preservation mechanisms. Such goals could be linked to ways of extracting metabolic energy to allow for such continued computation.
Background/professional context: artificial intelligence, artificial life, cognitive science, embodied cognition, enactivism, dynamical systems.
Coming from different fields and domains of practice, we found it challenging to converge on one definition of intelligence. Through hours of debate, conversation, and a fair amount of polite but energetic arguments, some shared aspects of what we each mean by intelligence started coming to light. Specifically, these were:
- a capacity for actively sensing and experiencing the world;
- behavior that is adaptive and learning in the world;
- an ability to construct, preserve and use memory;
- a way to extract meaning from the physical world into abstract structures and abstract it into another form;
- a process of understanding both permanent, relatively unchanging, qualities, as well as the fundamental impermanence of the world;
- a shared, collective, communal nature of cognition.
Moving from extremely concrete to ever more abstracted and general qualities of what we each understood by intelligence, we finally all agreed that in essence, in order to exhaustively address the issue of intelligence, we need to recognize that it is deeply linked to a more fundamental sense of being alive, to maintaining and preserving life.
Through these conversations, one particularly precious insight emerged: Our expressed disagreements reflect and problematize many of our own biases and implicit, tightly held beliefs. When these points of tension arise, a particular kind of intellectual, collective work becomes possible. This energy can be highly motivating and productive—worthy of noting as the first point in the methodological journaling of our project.
While in each of our respective fields we are certainly encouraged (and with good reason) to use well-defined terms, concretely embedded in our particular disciplinary history, and reflective of specific theoretical commitments, in this project we wish to keep alive a capacity to not quite know the right words to use. This allows us to retain strong links to the rigor of each discipline separately, while also creating space for our emerging, collective perspectives to be articulated and communicated back to the origin disciplines. We believe that differences are typically more productive than agreements, and the spirit of constructive, respectful and curious disagreement is clearly energising and will be key to the establishment of genuine interdisciplinary translation. Bringing distinct clusters of beliefs and expectations together around the concept of intelligence creates opportunities to explore the phenomena with a sharpened focus, born from tension.
Now, if by “intelligence” we mean all that was discussed above, what might we want to say is characteristic about the specific intelligent capacity that humans possess? Which would be the strengths and weaknesses of a distinctly human intelligence when compared to that of other intelligent agents, and in particular, AI? At our next meeting at CSAS we will address these questions and so take initial steps toward drawing a chart of intelligence as encountered in humans and machines. A report will follow here on the blog, and we warmly welcome your input.