Cognitive Computation: An Introduction
暂无分享,去创建一个
formulations (such as those of Aleksander). In the second keynote paper, McClelland shows how, even after more than half a century of research on machine intelligence, humans remain far better than the strongest computing machines available today at a wide range of natural cognitive tasks, such as object recognition, language comprehension, and planning and acting in contextually appropriate ways. After briefly reviewing the progress that is being made in many of these areas, he succinctly examines how and why computers still lack the fluidity, adaptability, open-endedness, creativity, purposefulness and insightfulness that are normally associated with the supreme achievements of human cognitive ability. Finally, he presents some exciting prospects for overcoming these limitations. Aleksander, in the third keynote paper, provides a comprehensive review of computational work that is currently developing under the heading of ‘Machine Consciousness’ and sets out to provide a guide for those who wish to contribute to this field. Initially, he raises and discusses questions of philosophical concern relating to the appropriateness of this activity and then describes a number of interesting classical designs and computational attitudes. This is followed by a convincing argument that shows that fine-grain neural approaches are needed to provide truly phenomenal representations that stand in relation to the behaviour of a computational organism, just as subjective mental states stand in relation to the existence of a conscious organism. He concludes the paper with an evaluation of the validity and benefits of designing conscious systems. In the next invited paper, Gurney makes an exciting and timely case for quantitative computational modelling as the only route to understanding cognition. Within this general strategy he argues that a programme of reverse engineering the brain, by building biologically constrained models using methods in computational neuroscience, holds most promise. In his ongoing attempts to address this grand challenge, the author outlines a four-level framework (computation, algorithm, mechanism and biological substrate) which provides a novel principled approach to model building. The author demonstrates the utility of the framework which can encompass working at multiple structural levels of description in the brain (from membranes to systems). Finally, the author describes a novel method involving the use of core-surround embedding for working at multiple levels simultaneously. Haikonen first reviews why the two traditional approaches towards artificial cognition, of symbolic artificial intelligence (AI) and sub-symbolic neural networks have not been very successful. He next shows how recent hybrid approaches that combine symbolic AI and sub-symbolic neural networks have also fallen short of the ultimate goal. The author argues that traditional AI programs do not operate with meanings and consequently do not understand anything. As a potential remedy, the author introduces and critically reviews the role of associative information processing principles for cognitive computing that may enable the utilization of meaning and the combined sub-symbolic/ symbolic operation of neural networks. Seth presents an excellent review of consciousness as a key feature of mammalian cognition. He reviews how computational and theoretical approaches can facilitate a transition from correlation to explanation in consciousness science. He succinctly describes progress towards identifying ‘explanatory correlates’ underlying a number of fundamental properties that characterize conscious experiences. He also discusses how synthetic approaches can shed additional light on possible functions of consciousness, the role of embodiment in consciousness and the plausibility of constructing a conscious artefact. Underwood presents a very interesting and timely review of models of attentional guidance in human image processing, with a focus on the visual saliency map hypothesis. His paper gives a ‘big picture’ perspective of how this work cumulates by evaluating the saliency map hypothesis of scene perception using evidence of eye movements made when images are first inspected. He concludes that visual saliency can be used by viewers, but that its use is both task-dependent and knowledgedependent. Gros addresses an important question in cognitive systems research, specifically of understanding the functional role of self-sustained neural activity in the brain and its interplay with the sensory data input stream. He reviews the present state of theoretical modelling and introduces an emerging approach to cognitive computation based on autonomously active neural networks. In contrast to the classical stimulus–response type neural networks, the author presents two novel neural architectures exhibiting continuous ongoing transient state dynamics in the context of a general critical discussion of the autonomous, self sustained activity of the brain. Sun presents a generic computational cognitive architecture emphasizing the role of motivational variables. The author convincingly argues that motivational representations can help make cognitive architectural models more comprehensive and provide deeper explanations of psychological processes. His pioneering work represents a step forward in making computational cognitive architectures better reflections of the human mind and its motivational complexity and intricacy. In the final paper of this Inaugural Issue, Ziemke and Lowe review the key role of emotion in embodied cognitive architectures. The authors succinctly argue that 2 Cogn Comput (2009) 1:1–3