Abstract Details

When are Atoms Happy?  Hartmut Neven , Peter Read (Visual Search, Google, Venice Beach, CA )   PL11

Any scientific attempt to explain consciousness is tasked with reconciling the third person perspective of science with our first person direct experience of the world. A good point of departure is to consider situations in which these two perspectives are correlated. We notice that behaviors conducive to our well being, i.e. conducive to maintaining homeostasis, tend to be associated with feelings of pleasure while actions that threaten our homeostasis tend to coincide with unpleasant feelings. We adopt a materialist/physicalist approach and offer two different pathways to explain this correlation, one explanation appeals to psychophysical parallelism while the other explanation invokes free will. Our considerations lead us to propose a three part design for an animat for which it can be argued that it possesses the analog of feelings. i) Drawing on insights gained from a recent experiment demonstrating quantum supremacy we propose a quantum circuit that prepares a state for which we argue it may be experienced as pleasurable by a system that assumes it. ii) Employing techniques from quantum error correction we implement a homeostatic loop allowing the system to protect the pleasant state against perturbations. We endow the system with the ability "to report feelings". We do so by constructing a mapping between the space of states seen by the interoception sensors of the animat which measure properties of the quantum state it seeks to stabilize and a space obtained from a word embedding of English language words describing feelings. iii) We integrate this homeostatic engine with a neural network such that incorrect classifications result in forces that threaten the stability of the pleasant state. We conjecture that a system concerned with maintaining homeostasis while learning is a better learner as compared to a system without a homeostatic engine.