Some issues concerning requirements for architectures, mechanisms, ontologies and forms of representation in intelligent human-like or animal-like robots are discussed. The tautology that a robot that acts and perceives in the world must be embodied is often combined with false premises, such as the premiss that a particular type of body is a requirement for intelligence, or for human intelligence, or the premiss that all cognition is concerned with sensorimotor interactions, or the premiss that all cognition is implemented in dynamical systems closely coupled with sensors and effectors. It is time to step back and ask what robotic research in the past decade has been ignoring. I shall try to identify some major research gaps by a combination of assembling requirements that have been largely ignored and design ideas that have not been investigated – partly because at present it is too difficult to make significant progress on those problems with physical robots, as too many different problems need to be solved simultaneously. In particular, the importance of studying some abstract features of the environment about which the animal or robot has to learn (extending ideas of J.J.Gibson) has not been widely appreciated. 1

This paper complements McCarthy’s “The well designed child”, in part by putting it in a broader context, the space of possible well designed progeny, and in part by relating design features to development of mathematical competence. I first moved into AI in an attempt to understand myself, especially hoping to understand how I could do mathematics. Over the ensuing four decades, my interactions with AI and other disciplines led to: design-based, cross-disciplinary investigations of requirements, especially those arising from interactions with a complex environment; a draft partial ontology for describing spaces of possible architectures, especially virtual machine architectures, for behaving systems (including our precursors); investigations of varied forms of representation and how they are suited to different functions; analysis of biological nature/nurture tradeoffs and their relevance to future machines; studies of control issues in a complex architecture; and showing how the states and processes possible in such an architecture relate to our (simplified) intuitive concepts of motivation, feeling, preferences, emotions, attitudes, values, moods, consciousness, etc. In 1971 I thought working models of human vision could lead to models of visual/spatial reasoning that would help to support Kant’s view of mathematics, against Hume’s. This has not yet happened, but I am still exploring

Some AI researchers aim to make useful machines, including robots. Others aim to understand general principles of information-processing machines with various kinds of intelligence, whether natural or artificial, including humans and human-like systems. They primarily address scientific and philosophical questions rather than practical goals. However, the tasks required to pursue scientific and engineering goals overlap, since both involve building working systems to test ideas and demonstrate results, and the conceptual frameworks and development tools needed for both overlap. This paper, partly based on philosophical analysis of requirements for robots in complex 3-D environments, surveys varieties of meta-cognition, drawing attention to requirements that drove biological evolution and which are also relevant to ambitious engineering goals.

This paper is an attempt to summarise and justify critical comments I have been making over several decades about research on consciousness by philosophers, scientists and engineers. This includes (a) explaining why the concept of “phenomenal consciousness” (P-C), in the sense defined by Ned Block, is semantically flawed and unsuitable as a target for scientific research or machine modelling, whereas something like the concept of “access consciousness” (A-C) with which it is often contrasted refers to phenomena that can be described and explained within a future scientific theory, and (b) explaining why the “hard problem ” is a bogus problem, because of its dependence on the P-C concept. It is compared with another bogus problem, “the ‘hard ’ problem of spatial identity ” introduced as part of a tutorial on semantically flawed concepts. Different types of semantic flaw and conceptual confusion not normally studied outside analytical philosophy are distinguished. The semantic flaws of the “zombie ” argument, closely allied with the P-C concept are also explained. These topics are related both to the evolution of human and animal minds and brains and to requirements for human-like robots. The diversity of the phenomena related to the concept “consciousness ” as ordinarily used makes it a polymorphic concept, partly analogous to concepts like “efficient”, “sensitive”, and “impediment” all of which need extra information to be provided before they can be applied to anything, and then

This chapter reports work done mostly by one member of the team – a philosopher with substantial AI programming experience, whose primary interests were in the very long term goals of the project, summarised in Chapter 1, including the goal of shedding light on problems solved by biological evolution,

Cognitive sensor networks are able to perceive, learn, reason and act by means of a distributed, sensor/actuator, computation and communication system. In animals, cognitive capabilities do not arise from a tabula rasa, but are due in large part to the intrinsic architecture (genetics) of the animal which has been evolved over a long period of time anddepends on a combination of constraints: e.g., ingest nutrients, avoid toxins, etc. We have previously shown how organism morphology arises from genetic algorithms responding to such constraints[6]. Recently, it has been suggested that abstract theories relevant to specific cognitive domains are likewise genetically coded in humans (e.g., language, physics of motion, logic, etc.); thus, these theories and models are abstracted from experience over time. We call this the Domain Theory Hypothesis, and other proponents include Chomsky [2] and Pinker [11] (universal language), Sloman [16, 17] (artificial intelligence), and Rosenberg [13] (cooperative behavior). Some advantages of such embedded theories are that they (1) make learning more efficient, (2) allow generalization across models, and (3) allow determination of true statements about the world beyond those available from direct experience. We have shown in previous work how theories of symmetry can dramatically improve representational efficiency and aid reinforcement learning on various problems [14]. However, it remains to be shown sensory data can be organized into appropriate elements so as to produce a model of a given theory. We address this here by showing how symmetric elements can be perceived by a sensor network and the role this plays in a cognitive system’s ability to discover knowledge about its own structure as well as about the surrounding physical world. Our view is that cognitive sensor networks which can learn these things will not need to be pre-programmed in detail for specific tasks. 1

1.2 A longer version of the story.................................. 3

is vision for, and how does it work?

• Very many researchers assume that it is obvious what vision (e.g. in humans) is for, i.e. what functions it has, leaving only the problem of explaining how those functions are fulfilled. • So they postulate mechanisms and try to show how those mechanisms can produce the required effects, and also, in some cases, try to show that those postulated mechanisms exist in humans and other animals and perform the postulated functions. • The main point of this presentation is that it is far from obvious what vision is for – and J.J. Gibson’s main achievement is drawing attention to some of the functions that other researchers had ignored. • I’ll present some of the other work, show how Gibson extends and improves it, and then point out much more there is to the functions of vision and other forms of perception than even Gibson had noticed. • In particular, much vision research, unlike Gibson, ignores vision’s function in on-line control and perception of continuous processes; and nearly all, including Gibson’s work, ignores meta-cognitive perception, and perception of possibilities and constraints on possibilities and the associated role of vision in reasoning.

What will it be like to admit Artificial Companions into our society? How will they change our relations with each other? How important will they be in the emotional and practical lives of their owners---since we know that people became emotionally dependent even on simple devices like the Tamagotchi? How much social life might they have in contacting each other? The contributors to this book assume that some form of long-term computer Companions are now a certainty in the coming years, and that it is a good moment to consider from a set of wide interdisciplinary perspectives, both how we shall construct them technically as well as their personal and social consequences. By Companions we mean conversationalists or confidants----not robots--- but rather computer software agents whose function will be to get to know their owners, who may well be elderly or lonely, and focusing not only on assistance via the internet (contacts, travel, doctors etc.) that many still find hard to use, but also on providing company and Companionship, by offering aspects of personalization. The human-Companion relationship could also be used to build a life narrative of the owner, eliciting over a long period a structure of the owner's life, perhaps in a level of detail that even their relatives might not recognize, or know about. You could call that autobiography