Mind-Brain Lecture: Adrian Cussins (Nat. Univ. of Colombia)
I would like to talk about a project ('CCP') that researches the logical origins of cognition and communication through designing, simulating and experimenting with a minimal world. It is not enough to simulate the 'mind half' of the mind-world interaction, since what matters for understanding the logical origins of cognition are the activity structures in virtue of which cognition is effective. These structures sometimes lean towards the mind side, and sometimes to the world side of the mind-world interaction. A modern human baby - even if its brain was a pure learning device - would not be a blank slate because it is born into an environment of highly sophisticated, established structures of language, cognition and culture. What, then, are the logical origins of these structures? CCP's minimal world begins without any conceptual cognitive structure, either in the mind or in the world. But only with the fact that the movement of its simple animals or 'bichos' leaves traces that are ontologically and cognitively consequential: these are traces of affect - affordance - solicitation. Out of these traces, worlds and minds are created. Telling the story of how is the goal of CCP. A guiding question for artificial intelligence from the 1950's was what is the simplest activity system which is such that it requires intelligence to perform well at it? This was AI's 'fruit-fly' question: as in molecular genetics, AI needs an object of study which is cheap, reproducible and simple but not so simple that it fails to manifest the phenomenon that is being investigated. Chess appeared to be a good candidate given the assumption that the essence of intelligence consists in the mechanisms whereby very large search spaces can be rendered tractable. Paradoxically, the success of DeepBlue showed the failure of that assumption. To understand what went wrong, and how to overcome it, we have developed a theory of the transition between a series of games: tic-tac-toe --> chess --> Go --> bagel-go (which is Go played on the surface of a torus, in order to avoid effects from the edges and corners of the board). CCP has been developing the next member in this series, which could be thought of as bagel-Go in which the 'players' live inside the game (so that all communication and understanding is part of the game-world, not transcendental to it); we call this 'yummyWorld', because it is based on the idea that the foundations or logical 'origins' of cognition and communication lie in a weird ontology of affect - affordance - solicitation. The basic form of perception is perception of affect (what is experienced as attractive / repulsive), a perception of affordance (what is experienced as offering a kind of action), and a kind of solicitation (experienced as demanding attention, or orienting attention, possibly as a kind of (mundane) normative call to the perceiver: 'here things go well / badly'). The simplest form of affect - affordance - solicitation is something like yummy and yucky. So yummyWorld has a board of 25 x 25 square grid arrayed over a torus, in which there are two 'teams' of four creatures each. When each creature moves, it leaves behind a trace which is experienced as yummy by its own team-mates and yucky by the others. Development of cognition and of the ontology of the environment proceeds in parallel, so there is no presupposition of a pre-formed world into which the creatures are born. The yummy / yucky traces function as both ontological and cognitive structures, which can develop into more complex structures with different properties: trails and territory. The creatures have to coordinate their movements through the environment in order to create territory. To this end they communicate, but what they can communicate is always at the same level as the structure of their environment-niche at the time of communication, and their cognitive state. One of the principles of yummyWorld is symmetry between perception, action, learning and communication: a creature (attempts to) move towards what it experiences as perceptually most salient, and it communicates what it experiences as perceptually most salient. The project also involves a novel system for learning in which there is *mundane* normative feedback; a notion influenced by the Chinese concept of Shi (something like, the dynamic potential born of the disposition and propensity of things) and notions of influence in Go. A creature must move at each game tick. A creature is drawn to yummy, and inhibited by yucky. If all around the creature (everything perceptible) is yucky then the creature cannot move, and so changes its affiliation from one team to the other. A team wins when all of the creatures belong to its affiliation. yummyWorld is a testbed with which to empirically develop and test theories of the logical origins of cognition and communication in kinds of content that are environmentally distributed, that presuppose no concepts, that belong neither to world nor to mind (or to both equally), and for which there is no distinction between content factors and motivational factors. Adrian Cussins is professor of philosophy at the National University of Colombia. His personal website is here: http://www.haecceia.com/FILES/research.htm.
