Towards living machines

Mechanical tortoise: A replica of W. Grey Walter’s 1951 robot Machina Speculatrix, incorporating several original parts, but without the perspex shell responsible for the tortoise nickname, built by Ian Horsfield and Owen Holland in the Bristol Robotics Lab. (Photo: Alan Winfield.) W. Grey Walter (1910–1977), a US-born neurophysiologist and EEG specialist at Bristol, UK, made important contributions to the neurosciences, but he is best remembered for a sideline activity, building robots. In an effort to show that extremely simple artificial brains can generate complex behaviour, Walter built two mechanical tortoises known as Elmer and Elsie in 1948–1949. The three-wheeled machines, which Walter called Machina Speculatrix, initially only had two ‘neurons’, enabling them to display phototactic behaviour. In later versions, Walter added a memory and a microphone and conducted Pavlovian experiments in which the machine was conditioned to associate specific sounds with rewards. As historian Andrew Pickering from the University of Exeter explained in the opening lecture of the recent Living Machines conference at the Natural History Museum in London, Walter’s tortoises hold an important position in the history of cybernetics, as they show the brain as an acting machine, not just a thinking machine. Efforts to imitate nature, Pickering elaborated, can produce very different results depending on which role of a biological system one decides to focus on. Mimicking only the thinking role of the brain, one can get artificial intelligence (AI) or chess-playing algorithms. Copying only mechanics gives us production-line robots and autonomous vacuum cleaners. However, if researchers can make artificial brains that actively engage with their environment, perceiving, thinking, and acting, then the result could be convincingly life-like, a true ‘living machine’.