As soon as we stop talking aloud, we seem to experience a kind of 'inner voice', a steady stream of verbal fragments expressing ongoing thoughts. What kind of information processing structures are required to explain such a phenomenon? Why would an 'inner voice' be useful? How could it have arisen? This paper explores these questions and reports briefly some computational experiments to help elucidate them.

There has been much discussion recently about the scope and limits of purely symbolic models of the mind and about the proper role of connectionism in cognitive modeling. This paper describes the symbol grounding problem : How can the semantic interpretation of a formal symbol system be made intrinsic to the system, rather than just parasitic on the meanings in our heads? How can the meanings of the meaningless symbol tokens, manipulated solely on the basis of their shapes, be grounded (...) in anything but other meaningless symbols? The problem is analogous to trying to learn Chinese from a Chinese/Chinese dictionary alone. A candidate solution is sketched: Symbolic representations must be grounded bottom-up in nonsymbolic representations of two kinds: iconic representations, which are analogs of the proximal sensory projections of distal objects and events, and categorical representations, which are learned and innate feature-detectors that pick out the invariant features of object and event categories from their sensory projections. Elementary symbols are the names of these object and event categories, assigned on the basis of their categorical representations. Higher-order symbolic representations, grounded in these elementary symbols, consist of symbol strings describing category membership relations. Connectionism is one natural candidate for the mechanism that learns the invariant features underlying categorical representations, thereby connecting names to the proximal projections of the distal objects they stand for. In this way connectionism can be seen as a complementary component in a hybrid nonsymbolic/symbolic model of the mind, rather than a rival to purely symbolic modeling. Such a hybrid model would not have an autonomous symbolic module, however; the symbolic functions would emerge as an intrinsically dedicated symbol system as a consequence of the bottom-up grounding of categories ' names in their sensory representations. Symbol manipulation would be governed not just by the arbitrary shapes of the symbol tokens, but by the nonarbitrary shapes of the icons and category invariants in which they are grounded. (shrink)

The first brief description is given of a project aimed at searching for the neural correlates of consciousness through computer simulation. The underlying model is based on the known circuitry of the mammalian nervous system, the neuronal groups of which are approximated as binary composite units. The simulated nervous system includes just two senses - hearing and touch - and it drives a set of muscles that serve vocalisation, feeding and bladder control. These functions were chosen because of their relevance (...) to the earliest stages of human life, and the simulation has been given the name CyberChild. The system's pain receptors respond to a sufficiently low milk level in the stomach, if there is simultaneously a low level of blood sugar, and also to a full bladder and an unchanged diaper. It is believed that it may be possible to infer the presence of consciousness in the simulation through observations of CyberChild's behaviour, and from the monitoring of its ability to ontogenetically acquire novel reflexes. The author has suggested that this ability is the crucial evolutionary advantage of possessing consciousness. The project is still in its very early stages, and although no suggestion of consciousness has yet emerged, there appears to be no fundamental reason why consciousness could not ultimately develop and be observed. (shrink)

What psychological and philosophical significance should we attach to recent efforts at computer simulations of human cognitive capacities? In answering this question, I find it useful to distinguish what I will call "strong" AI from "weak" or "cautious" AI. According to weak AI, the principal value of the computer in the study of the mind is that it gives us a very powerful tool. For example, it enables us to formulate and test hypotheses in a more rigorous and precise fashion. (...) But according to strong AI, the computer is not merely a tool in the study of the mind; rather, the appropriately programmed computer really is a mind, in the sense that computers given the right programs can be literally said to. (shrink)

I have argued in a number of writings[1] that the philosophical part (though not the neurobiological part) of the traditional mind-body problem has a fairly simple and obvious solution: All of our mental phenomena are caused by lower level neuronal processes in the brain and are themselves realized in the brain as higher level, or system, features. The form of causation is.

This paper responds to the issues raised by D. Chalmers by offering a research direction which is quite radical because of the way in which methodological principles are linked to scientific studies of consciousness. Neuro-phenomenology is the name I use here to designate a quest to marry modern cognitive science and a disciplined approach to human experience, thereby placing myself in the lineage of the continental tradition of Phenomenology. My claim is that the so-called hard problem that animates these Special (...) Issues can only be addressed productively by gathering a research community armed with new pragmatic tools for the development of a science of consciousness. I will claim that no piecemeal empirical correlates, nor purely theoretical principles, will really help us at this stage. We need to turn to a systematic exploration of the only link between mind and consciousness that seems both obvious and natural: the structure of human experience itself. In what follows I motivate my choice by briefly examining the current debate about consciousness at the light of Chalmer’s hard problem. Next, I outline the phenomenological strategy. Finally I conclude by discussing some of the main difficulties and consequences of this strategy. (shrink)

We are engineers, and our view of consciousness is shaped by an engineering ambition: we would like to build a conscious machine. We begin by acknowledging that we may be a little disadvantaged, in that consciousness studies do not form part of the engineering curriculum, and so we may be starting from a position of considerable ignorance as regards the study of consciousness itself. In practice, however, this may not set us back very far; almost a decade ago, Crick wrote: (...) 'Everyone has a rough idea of what is meant by consciousness. It is better to avoid a precise definition of consciousness because of the dangers of premature definition. Until the problem is understood much better, any attempt at a formal definition is likely to be either misleading or overly restrictive, or both'. This seems to be as true now as it was then, although the identification of different aspects of consciousness by Block has certainly brought a degree of clarification. On the other hand, there is little doubt that consciousness does seem to be something to do with the operation of a sophisticated control system, and we can claim more familiarity with control systems than can most philosophers, so perhaps we can make up some ground there. (shrink)

This paper relates to a formal statement of the mechanisms that are thought minimally necessary to underpin consciousness. This is expressed in the form of axioms. We deem this to be useful if there is ever to be clarity in answering questions about whether this or the other organism is or is not conscious. As usual, axioms are ways of making formal statements of intuitive beliefs and looking, again formally, at the consequences of such beliefs. The use of this style (...) of exposition does not entail a claim to provide a mathematically rigorous formal deductive system. Conventional mathematical notation is used to achieve clarity, although this is elaborated with natural language in an attempt to reduce terseness. In our view, making the approach axiomatic is synonymous with building clear usable tests for consciousness and is therefore a central feature of the paper. The extended scope of this approach is to lay down some essential properties that should be considered when designing machines that could be said to be conscious. In the broader discussion about the nature of consciousness and its neurological mechanisms, it may seem to some that axiomatisation is premature and continues to beg many questions. However, the approach is meant to be open- ended so that others can build further axiomatic clarifications that address the very large number of questions which, in the search for a formal basis for consciousness, still remain to be answered. Of course, in discussions about consciousness many will also argue that the subject is not one that may ever be formally addressed by means of axioms. The view taken in this paper is 'let's try to do it and see how far it gets'. (shrink)

What type of artificial systems will claim to be conscious and will claim to experience qualia? The ability to comment upon physical states of a brain-like dynamical system coupled with its environment seems to be sufficient to make claims. The flow of internal states in such system, guided and limited by associative memory, is similar to the stream of consciousness. Minimal requirements for an artificial system that will claim to be conscious were given in form of specific architecture named articon. (...) Nonverbal discrimination of the working memory states of the articon gives it the ability to experience different qualities of internal states. Analysis of the inner state flows of such a system during typical behavioral process shows that qualia are inseparable from perception and action. The role of consciousness in learning of skills, when conscious information processing is replaced by subconscious, is elucidated. Arguments confirming that phenomenal experience is a result of cognitive processes are presented. Possible philosophical objections based on the Chinese room and other arguments are discussed, but they are insufficient to refute claims articon’s claims. Conditions for genuine understanding that go beyond the Turing test are presented. Articons may fulfill such conditions and in principle the structure of their experiences may be arbitrarily close to human. (shrink)

A "machine" is any causal physical system, hence we are machines, hence machines can be conscious. The question is: which kinds of machines can be conscious? Chances are that robots that can pass the Turing Test -- completely indistinguishable from us in their behavioral capacities -- can be conscious (i.e. feel), but we can never be sure (because of the "other-minds" problem). And we can never know HOW they have minds, because of the "mind/body" problem. We can only know how (...) they pass the Turing Test, but not how, why or whether that makes them feel. (shrink)

This paper proposes a brain-inspired cognitive architecture that incorporates approximations to the concepts of consciousness, imagination, and emotion. To emulate the empirically established cognitive efficacy of conscious as opposed to non-conscious information processing in the mammalian brain, the architecture adopts a model of information flow from global workspace theory. Cognitive functions such as anticipation and planning are realised through internal simulation of interaction with the environment. Action selection, in both actual and internally simulated interaction with the environment, is mediated by (...) affect. An implementation of the architecture is described which is based on weightless neurons and is used to control a simulated robot. (shrink)

To explore issues of developmental structure, physical em- bodiment, integration of multiple sensory and motor systems, and social interaction, we have constructed an upper-torso humanoid robot called Cog. The robot has twenty-one degrees of freedom and a variety of sen- sory systems, including visual, auditory, vestibular, kinesthetic, and tac- tile senses. This chapter gives a background on the methodology that we have used in our investigations, highlights the research issues that have been raised during this project, and provides a summary (...) of both the current state of the project and our long-term goals. We report on a variety of implemented visual-motor routines (smooth-pursuit track- ing, saccades, binocular vergence, and vestibular-ocular and opto-kinetic re?exes), orientation behaviors, motor control techniques, and social be- haviors (pointing to a visual target, recognizing joint attention through face and eye ?nding, imitation of head nods, and regulating interaction through expressive feedback). We further outline a number of areas for future research that will be necessary to build a complete embodied sys- tem. (shrink)

This paper presents a computer model of cortical broadcast and competition based on spiking neurons and inspired by the hypothesis of a global neuronal workspace underlying conscious information processing in the human brain. In the model, the hypothesised workspace is realised by a collection of recurrently inter-connected regions capable of sustaining and disseminating a reverberating spatial pattern of activation. At the same time, the workspace remains susceptible to new patterns arriving from outlying cortical populations. Competition among these cortical populations for (...) influence on the workspace is effected by a combination of mutual inhibition and top-down amplification. (shrink)

Conscious events interact with memory systems in learning, rehearsal and retrieval (Ebbinghaus 1885/1964; Tulving 1985). Here we present hypotheses that arise from the IDA computional model (Franklin,Kelemen and McCauley 1998; Franklin 2001b) of global workspace theory (Baars 1988, 2002). Our primary tool for this exploration is a flexible cognitive cycle employed by the IDA computational model and hypothesized to be a basic element of human cognitive processing. Since cognitive cycles are hypothesized to occur five to tentimes a second and include (...) interaction between conscious contents and several of the memory systems, they provide the means for an exceptionally fine-grained analysis of various cognitive tasks. We apply this tool to the small effect size of subliminal learning compared to supraliminal learning, to process dissociation, to implicit learning, to recognition vs. recall, and to the availability heuristic in recall. The IDA model elucidates the role of consciousness in the updating of perceptual memory, transient episodic memory, and procedural memory. In most cases, memory is hypothesized to interact with conscious events for its normal functioning. The methodology of the paper is unusual in that the hypotheses and explanations presented are derived from an empirically based, but broad and qualitative computational model of human cognition. (shrink)

Many materialists believe that we should, in principle, be able to build a conscious computing machine. Others disagree. I favour a sceptical position, but of another variety. The problem isn't that it would be impossible to create a conscious computer. The problem is that we cannot know whether it is possible. There are principled reasons for thinking that we wouldn't ever be able to confirm that allegedly conscious computers were conscious. The proper stance on computational consciousness is agnosticism. Despite this (...) agnosticism, I think we are very close to understanding the material basis of consciousness. Close, but we will never get all the way there. Our understanding of the material basis of consciousness is ineluctably incomplete. That makes me a mysterian. But I am not a defeatist mysterian. I do not think the irresolvable mysteries of consciousness prevent us from formulating concrete empirically grounded theories of consciousness. I will even outline such a theory below. I also think we can identify some properties that are necessary for consciousness, and some that are sufficient. The problem is we cannot find properties that are both. This may sound like a contradiction, but it is not. We can know necessary properties and sufficient properties without knowing necessary and sufficient properties. One can know that drinking two litres of alcohol is sufficient for intoxication, and that at least one teaspoon of alcohol is necessary without knowing the quantity that is both necessary and sufficient. In the case of consciousness, the properties we know to be sufficient are properties that computers lack. The properties we know to be necessary can be possessed by computers, but we don't know whether those properties suffice. We can equip computers with every knowable necessary property and still scratch our heads when asked whether we have managed to create artificial experience. (shrink)

1 INSERM-CEA Unit 562, Cognitive Neuroimaging, Service Hospitalier Fre´de´ric Joliot, Orsay, France, 2 CNRS URA2182 Re´cepteurs and Cognition, Institut Pasteur, Paris, France.

This paper looks closely at previously enunciated axioms that specifically include phenomenology as the sense of a self in a perceptual world. This, we suggest, is an appropriate way of doing science on a first-person phenomenon. The axioms break consciousness down into five key components: presence, imagination, attention, volition and emotions. The paper examines anew the mechanism of each and how they interact to give a single sensation. An abstract architecture, the Kernel Architecture, is introduced as a starting point for (...) building computational models. The thrust of the paper is to relate the axioms to the kernel architecture and indicate that this opens a way of discussing some first-person issues: tests for consciousness, animal consciousness and Higher Order Thought. (shrink)

A connectionist system that is capable of learning about the spatial structure of a simple world is used for the purposes of synthetic epistemology: the creation and analysis of artificial systems in order to clarify philosophical issues that arise in the explanation of how agents, both natural and artificial, represent the world. In this case, the issues to be clarified focus on the content of representational states that exist prior to a fully objective understanding of a spatial domain. In particular, (...) the criticisms of (Chrisley, 1993) that were raised in (Holland, 1994) are addressed: how can we determine that a system’s spatial representations are more objective than before? And under what conditions (tasks, training regimes, environments) do such increases in objectivity occur? After analysing the results of experiments that attempt to shed light on these questions, the study concludes by comparing and contrasting this work with related research. (shrink)