Are robots smart?

Citation

Parker, G. (2001), "Are robots smart?", Industrial Robot, Vol. 28 No. 4. https://doi.org/10.1108/ir.2001.04928daa.002

Are robots smart?

Are robots smart?

Keywords: Robots, Artificial intelligence

Academics get asked from time to time to give talks to sixth-form students in schools to generate interest in engineering and science. I am no exception, and started to think about popular misconceptions about robots and their real or potential impact on society. Most people think of robots in humanoid form generated from their contact with science-fiction films. Daleks ran for a long time in the very popular BBC series Dr Who. So do we think from this that such machines are smart or "intelligent" and will ultimately rule our lives?

If we go back to machine basics, the advent of the microprocessor allowed complex articulations, such as a robot arm, to perform forward and inverse transformations in real time between the robot and real-world co-ordinate systems. Software command structures then provided for such robots to be programmed, usually by teach mode, to carry out a variety of complex, but repetitive, tasks. Thus was born the industrial robot, which continues to expand impressively into a wide variety of manufacturing and other applications. World-wide sales are now reaching about 700,000. However, despite the increasing sophistication of such machines with integrated sensor technology, for example vision systems, they could not be classified as particularly smart. They are, in effect, extensions of existing flexible automation and, despite their large numbers, are little known by the general public.

In looking for progression towards an intelligent machine concept one can think of two distinct routes. First, the remote, teleoperated robot and, second, the learning robot. In the first category there are numerous examples of the effectiveness of robots operating in hazardous or remote environments. Popular robotics on television apparently shows Star Wars as sophisticated warring between rival machines, but the reality is radio-controlled vehicles operated by humans. More sophisticated examples include nuclear power plants, submersible craft, Mars surface robots, medical surgery, bomb disposal, and so on, which all illustrate the powerful combination of various robotic devices controlled remotely by human beings. The intelligence of the human is combined with a device well-suited for the environment to provide a solution not possible with an entirely intelligent autonomous robot solution. Machines are capable of storing vast amounts of information and are good at multi-tasking, whereas the human has sophisticated perception and cognition. We see that by harnessing the human and the machine in co-operation sophisticated tasks may be undertaken.

Despite this, further advances are being made in carrying out tasks by the use of augmented reality. This involves superimposing graphical overlays on real images of the remote scene so as to convey extra information. If the cameras are calibrated it is possible to register the two scenes and convey quantitative information between the real and graphical worlds. For example, my research group has successfully used this technique for controlling a remote camera mounted robot inspecting the condition of sewer pipes.

The second route towards the intelligent robot involves the development of learning software, mainly based on neural network principles, to teach a robot to undertake new tasks which it has not achieved before. Artificial neural networks resemble human neural networks but are far simpler and rely on being taught a set of input data related to the problem. Proponents claim that it is only a matter of time before such networks take on human characteristics. For example, the latest Japanese walking robot, Pino, has neural network circuits which, it is claimed, will one day mimic the human brain. Recently, a Russian scientist, Vitali Valtsev, has created a neuro-computer from neural networks which he claims is able to make judgements according to the circumstances and learn from mistakes. Despite these advances there are questions as to whether genuine advances in some areas of artificial intelligence have really been made; such as in decision making and reasoning. Can we also conceive a machine which can function in an environment completely outside that which has been defined by the human designer?

So, at the present time remote, teleoperated robots look quite smart thanks to their human operator, but in the longer term learning robots may do better despite some doubts about intelligence. Then we might see a genuine robot wars contest!

Graham Parker