BlueSci Issue 02 - Lent 2005

Page 13

Robots: the Next Generation? Anna Lacey goes in search of artificial life The Terminator, C-3PO and HAL: Hollywood’s robot stars. Scientists and engineers have spent years attempting to turn this science fiction into reality, but have we really come any closer to creating an artificially intelligent being? Amateur scientist and self-proclaimed “nerd with a mission”, Steve Grand has come closer than some. Close enough, in fact, for Richard Dawkins to label him “the creator of what I think is the nearest approach to artificial life so far”. Grand’s aim is to create a machine ‘brain’; a machine capable of self-organising into a series of more specialised machines, with only sensory information as a guide. Lucy the orang-utan is the first step on his mammoth quest.

Lucy the orang-utan is the first step on his mammoth quest

Lucy (pictured on the right) comes fully equipped with binaural hearing, monocular vision, virtual muscles, touch and temperature sensors, and even a voice. Information from the environment is detected by these sense devices and passed to her brain. The brain itself consists of over 50, 000 virtual neurons, which work together to form an array of neural circuits. Grand aims to show that these circuits can interact to produce outcomes that are unpredictable. In other words, that complexity can arise from apparently simple beginnings.With this intellectual toolbox behind her, Lucy can distinguish a banana from an apple, a simple feat for a human, but impressive for a robotic orangutan. How does she do this? The answer lies in the building of Lucy’s virtual brain. Grand used conventional computer programmes to simulate different brain structures.This included a model of the superior colliculus, the part of the brain that receives visual information and stimulates motor responses.The visual part of Lucy’s ‘brain’ enables her to differentiate between apples and bananas, while the motor part allows her to move her eyes and arms to a visual point in space. But why would Lucy want to point at a banana rather than an apple? It is certainly not due to Lucy having a penchant for tropical fruit. Instead, Grand has had to build a preference for bananas into her program, as the reasoning behind why organisms choose one option over another is far beyond our current understanding.The programming integrates to allow Lucy to distinguish boundaries, recognise two fruit shapes, realise her intrinsic preference for the long, yellow fruit and final-

www.bluesci.org

Steve Grand

ly point at the banana. Joining together these fairly simple components makes Lucy special: separate areas of her ‘brain’ are programmed with individual functions, and yet together they interact to combine visual sensation with movement. In many ways this does not seem like ground-breaking robotics. After all, there are many pre-programmed house robots that can sense the environment and respond accordingly. Even Grand admits that Lucy only points at bananas because this is what she was wired up to do. Despite

Lucy’s true importance lies in her giving us a tantalising glimpse of how we think

this, Grand’s virtual orang-utan still deserves a place in history. Although Lucy was given pre-programmed machinery, she didn’t actually know anything about apples and bananas, and so had to work it out for herself. Grand argues that this shows neural circuits organising themselves into something more complex, and is thus an example of development and learning.

The difficulty arises in knowing exactly what it is ‘to learn’. Definitions and mechanisms of learning and intelligence have long been debated by philosophers and scientists. Without understanding what it means to say ‘I learn’ on a descriptive or mechanical level, it is impossible to judge whether Lucy has learnt or not. Lucy’s true importance lies in her giving us a tantalising glimpse of how we think. If simple circuits can allow a robot to recognise a banana and point at it, then there is no reason to think that our own brain cells cannot create more complicated networks and outcomes. So should the Cambridge applicants of the future worry about competing for places with robots? Grand thinks not. If a real breakthrough in artificial intelligence is to be made, scientists must work out the processes behind learning, intelligence and creativity. “Until then, there’s the small matter of making Lucy smart enough to pick up the application form!” says Grand. Understanding the brain through experimenting with circuits seems a worthwhile course for now.After all, evolution did not get the wiring perfect first time. Perhaps one day an amateur in a shed will strike lucky and get it right. Anna Lacey is a third year Natural Scientist specialising in Zoology

11


Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.