| | A World of Human-like Machines | The science of robotics, which draws on other disciplines such as artificial intelligence and micro-engineering, is generally understood to concern the design of autonomous or semi-autonomous machines, often modelled directly on human attributes and skills. The military have shown a particular interest in automated weaponry and mechanically intelligent surveillance devices, for obvious reasons, and it is certainly the case that a large proportion of current research projects are funded directly or indirectly by the US agency DARPA (Defence Advanced Research Projects Agency). Manuel De Landa (1991) has effectively portrayed the historical precedents and potentially disturbing consequences of automated war in War in the Age of Intelligent Machines. He argues the twentieth century saw a shift in the relation between humans and machines that may lead eventually to the emergence of a truly independent robotic life-form, a “machinic phylum” to use a phrase he borrows from Gilles Deleuze. Meanwhile, advances in computer control through parallel processing and learning systems that produce semi-intelligent robots, or ‘knowbots’, have accelerated the integration of machines into mass production. Here productivity is increased and labour costs reduced by the automation of many processes leading to a situation where manufacturing lines are often human-free zones as many tasks that previously required great human skill and dexterity are mechanised. And while industrial robots are now relatively static and cumbersome, the aim of much current robotic research is to achieve autonomy for the machine, to free it from static sources of power and human intervention. Mobile robots, or ‘mobots’, are intended for applications in space exploration, warfare and nuclear installations but may eventually find their way into the home in domestic applications. Most robots in use today are blindly pre-programmed to do repetitive tasks, but research into machine vision, sound sensing and touch sensitivity will allow them to sense their environment and take ‘real-time’ decisions about their operation. At the same time as investments are made in large-scale robotic projects, alternative methods are explored that distribute resources rather than concentrating them. Rodney Brooks at the Massachusetts Institute of Technology (MIT) has proposed robots that are “Fast, Cheap and Out of Control”, consisting of millions of tiny units, each programmed to do a simple task, but not subject to any centralised control. In this sense they are like an ant colony that can build large structures through the co-operation of lots of tiny workers. Brooks suggests that such creatures could be dropped on a planet surface and work together to clear an area of rocks for a landing pad. It would not matter that many of the minibots might die or stop working, because they can easily be replaced. This is an example of human engineering trying to model technology from nature to improve efficiency. Equally interesting is the seemingly awesome power of Mark Tilden’s ‘Unibug’ made from cast-off electrical parts assembled for a couple of hundred dollars and described in Robosapiens. The Unibug, almost uniquely amongst current robots, dispenses with digital processing and uses analogue feedback circuits which allow this little ‘creature’ to move about and learn. These units are highly efficient, very cheap and more reliable than many more expensive systems. At the other end of the complexity spectrum, Rodney Brooks has recently suggested that humans and machines will shortly reach a level of equivalent intelligence and worldly behaviour, and that we will increasingly come to see robots as companions and guides. The dream of creating intelligent mechanical objects has historically been bound up with the strong AI (artificial intelligence) goal of modelling the human brain in order to replicate the mind. However… traditionally this has tended towards a rather ‘disembodied’ understanding of the mind as a ‘brain-determined’ phenomenon. Taking their cue from the ‘situatedness’ of the embodied human brain, a new generation of researchers are building systems that more closely mimic the real behaviour of brains and bodies in the world by combining AI and robotic systems. This kind of work is being conducted using a $1 million ‘Dynamic Brain’ robot at the Japanese ATR Centre just outside Tokyo under the direction of Stephan Shaal and Mitsuo Kawato. But despite all the excitement and the high expectations of robotics it should also be recognised that we are still coming to terms with the huge degree of complexity involved in replicating anything approaching human-like behavior (or ‘humanoid’ as the terminology has it). Even given the remarkable balance and agility of the Honda Corporation’s hugely expensive ‘Humanoid Robot’ (http://world.honda.com/robot/) and its ability to walk down stairs and kick a ball, you probably wouldn’t trust it to wash your best wine glasses. There is a danger that high-end robotic research comes to be seen as a public-relations exercise for large businesses, with few practical applications. In response, funding-hungry research is setting its sights on smaller, more achievable, areas of investigation such as ‘search and rescue’ and surgical assistance where practical benefit can more readily accrue by extending human abilities rather than replicating them. So while theorists and designers like Rodney Brooks, Ray Kurzweil and Hans Moravec are confidently predicting humanoid beings within the century, it is clear that the compelling vision for those leading the field is of a world co-inhabited by human-like machines. | — Robert Pepperell, The Posthuman Condition - Consciousness Beyond the Brain | Indexes/23 |
0 comments:
Post a Comment