We present a position paper of work-in-progress on homeostatically motivated artificial intelligence.
We propose the design and construction of a new class of machines organized according to the principles of life regulation. These machines have physical constructions— bodies—that must be maintained by homeostasis within a narrow range of viability states and thus share some essential traits with all living systems. The fundamental innovation of these machines is the introduction of risk to self. Rather than up-armoring or adding raw processing power to achieve resilience, we begin the design of these robots by, paradoxically, introducing vulnerability. Drawing on recent developments in soft robotics and algorithms for multisensory abstraction, we outline a way forward to more biologically authentic implementations of feeling in machines.
The main motivation for this project is a set of theoretical contributions to the
understanding of biological systems endowed with feeling. Damasio has provided a
rationale for the emergence of feelings as a means to optimize survival, a process that is
closely associated with the biological basis of selfhood
Machines that implement a process resembling homeostasis might acquire a new source of motivation and evaluation of behavior, akin to that of feelings in living organisms. The world affords risks and opportunities, not in relation to an arbitrary reward or loss function, but in relation to the continued existence of the robot itself. Rewards aren’t rewarding and losses don’t hurt unless they are rooted in the ground of life and death. True agency arises when the machine picks a side in this dichotomy, when it acts with a preference for existence over dissolution. A robot engineered to participate in its own homeostasis might finally become its own locus of concern. This elementary
concern infuses meaning into information processing. A robot operating on intrinsically meaningful representations might seek broadly intelligent solutions to the human relevant tasks set before it, natural and social.
Our approach diverges from traditional conceptions of intelligence that emphasize outward-directed perception and abstract problem solving. We regard high-level cognition as an outgrowth of resources that originated to solve the ancient biological problem of homeostasis. A physical body is subject to perennial risk and decay. But to a disembodied algorithm—or to a robot whose physical persistence is guaranteed—nothing can happen that carries any consequences. Sense-data become meaningful when the data are connected to the maintenance and integrity of the sensing agent and to its homeostasis. Sensory processing that is not attached to a vulnerable body “makes no sense”.
The presence of a body serving as an aid or scaffold to problem-solving does not
suffice to generate meaning – otherwise all physical robots would be trivially eligible.
Preliminary steps have been taken to add “emotional circuits” to influence robotic goal
selection
Di Paolo (2003) has criticized the program of embodied robotics as still missing an
organism-level logic. Previous work has produced agents that can regulate their own
susceptibility to environmental cues based on abstract internal variables
We advocate a transition from embodied artificial intelligence to homeostatically
motivated artificial intelligence. Intelligence has been defined as “an agent’s ability to
achieve goals in a wide range of environments.”
The past decade has witnessed the birth of the sub-discipline of soft robotics. This was enabled by new discoveries in the design and construction of soft “tissues” embedded with electronics, sensors, and actuators. These artificial tissues are flexible, stretchable, compressible, and bounce back resiliently. In short, they are naturally compliant with their environments. Soft materials potentially provide a richer source of information on body and environment than do traditional rigid materials.
Living organisms are composed of tissues that participate in their own self-maintenance. Living tissues sense and signal the state of their life process. In contrast, consider the “life” of a piece of sheet metal bent into a boxy robot. In general, metal is so durable in comparison to its niche – our human environment – that its security and viability are afterthoughts. Metals and hard plastics are ubiquitous in robotics with the consequence that, in most cases, material integrity can be safely ignored. But durability comes at a cost. An invulnerable material has not much to say about its well-being. If we imagine strain gauges embedded throughout a hard surface, they would spend most of their time reporting “no change”, until a catastrophic failure occurs and all the sensors cry out for the first and last time.
Soft robots, on the other hand, more readily enter into a graceful and sensitive
coupling with the environment
The effort to build a robot with a homeostatic self representation would present a
complex exercise in machine learning. It could draw upon a neuroarchitectural framework
originally proposed in 1989
Crucially, crossmodal processing can accommodate the interoceptive modalities. A
proposal has been made that the feeling of existence itself, or of conscious presence,
may be due to predictive coding of internal sensations
There is an intriguing correspondence between the biological architecture of sensory convergence and some variants of deep neural networks. We suggest that artificial neural networks are poised to tackle the next great challenge of building correspondences between inner space and outer space, between internal homeostatic data and external sense data.
A machine in charge of its own self-regulation, and constructed of soft and sensitive tissues, will have a wealth of internal data upon which to draw to inform its plans and perceptions. The homeostatic robot will process information with the aid of something akin to feeling. How does the color, taste, and texture, say, of an apple, systematically associate with changes to the ongoing management of life? All of which is to say that the question “How does this make you feel?” might soon be asked of machines.