ANALOG VLSI VISION SYSTEMS Vision is the process of extracting high level information from two dimensional data arrays produced by energy (frequently visible light) reflected or transmitted from objects in the real world. Human vision is a complex process which builds a coherent three dimensional, realtime model of the world from scenes of arbitrary complexity. The first stages of the human computer vision process occurs in the first few layers of the retina. My current project is an silicon retina using mixed analog and digital CMOS technology with a base feature size (lambda) of .18 microns. This chip can perform arbitrary convolution operations (operations which produce an output array based on the immediate neighbourhood of each input array cell) using analog product and sum circuits. These operations can be used to perform some of the early vision processing done in biological retinas. The advantages of this approach include extremely high effective processing speeds and the very low costs of a CMOS solution, at the cost of a lower signal to noise ratio compared with other imaging array types (CCD). By using the advantages of high speed signal processing to compensate for the lower signal to noise ratio a produce a low cost, high capability smart vision chip. ARTIFICIAL LIFE Artificial Life is a broad field encompassing the study of artificial systems which in some way mimic biological organisms. My research in this area focuses on the evolution of co-operative behaviours in organisms which are able to estimate their degree of relatedness with other organisms. Non-reciprocal altruism is able to evolve under conditions where the altruist and the recipient are related, and where the cost to the altruist is less than the benefit accrued to the recipient times the degree of relatedness between the altruist and the recipient (for example a parent/child or sibling relationship implies a relatedness coefficient of .5). Evolutionary costs and benefits are indirectly measured in terms of ultimate reproductive success. Because the true index of relatedness is not directly available to organisms they must use other markers to decide whether or not to make an altruistic gift to another organism. In experiments, organisms capable of kin recognition demonstrate tremendous evolutionary advantages over organisms which are not. This has obvious analogs to real world biological systems and may prove useful in developing useful software agents which cooperate to solve group tasks.