• We often don’t notice it, but artificial intelligence (AI) is all around us. It is present in computer games, in the cruise control in our cars and the servers that route our email.
• The world’s most powerful supercomputer can carry out 100 trillion operations per second, which some scientists believe could be approaching the processing power of the human brain.
• Machines have always excelled at tasks like calculation. But they can now challenge humans on everything from chess to football to mixing music.
What is AI?
So, can a machine behave like a person? This question underlies artificial intelligence (AI), the study of intelligent behaviour in machines. In the 1980s, AI research focused on creating machines that could solve problems and reason like humans.
One of the most difficult problems in artificial intelligence is that of consciousness. A consciousness gives us feelings and makes us aware of our own existence. But scientists have found it difficult getting robots to carry out even the simplest of cognitive tasks.
Creating a self-aware robot with real feelings is a significant challenge faced by scientists hoping to mimic human intelligence in a machine.
Since the early 1990s, researchers have concentrated on developing smaller, independent robots instead of trying to recreate human intelligence. The model for many of these machines is insect intelligence, which is – in its own way – very sophisticated.
When it is completed in late 2004, the world’s most powerful computer will be ASCI Purple, built by IBM. It is expected to carry out 100 trillion operations per second (or 100 teraflops). A supercomputer with double this processing power is expected within the next two years. It is being built to replace ASCI White – formerly the world’s most powerful computer – which occupies a space the size of two basketball courts at the Lawrence Livermore National Laboratory in Livermore, California.
A spokesman for IBM said that ASCI Purple was approaching the processing power of the human brain. But some scientists believe our brains can carry out around 10,000 trillion operations per second. HAL, the supercomputer that rebels against its human handlers in the film 2001: A Space Odyssey (1968), is a cheeky reference to IBM. The letters H, A and L, precede I, B and M in the alphabet.
In 1950, mathematician Alan Turing devised a test to identify whether a machine displayed intelligence. In the Turing Test, two people (A and B) sit in a closed room, while an interrogator (C) sits outside. Person A tries to fool the interrogator about their gender, while person B tries to assist the interrogator in their identification. Turing suggested a machine take the place of person A. If the machine consistently fooled the human interrogator, it was likely to be intelligent.
The possible dangers posed by intelligent machines have inspired countless science fiction films. In The Terminator (1984), a computer network nukes the human race in order to achieve supremacy. This network then manufactures intelligent robots called ‘Terminators’ which it programs to annihilate human survivors.
In The Matrix (1999) and The Matrix Reloaded (2003), a machine enslaves humanity, using people as batteries to power its mainframe. Steven Spielberg’s AI: Artificial Intelligence (2002) paints a more sympathetic view of artificial life, depicting sensitive robots that are abused by brutal, selfish human masters.
Man and machine
One place where artificial intelligence has found a natural home is in the development of computer games. AI in computer games is becoming increasingly sophisticated as consumer appetites for better, faster, more challenging games grows. In games, AI is often present in the opponents you play against, or in allies or other team members.
In the RoboCup junior championships, independent robots compete one-on-one on a miniature football table with a greyscale pattern. The robots use this dark to pale gradient to navigate their way to the opponent’s goal. A special ball is used which contains sensors that communicate with sensors in the robot.
Despite these entertaining applications, the original point of AI research was to create machines that could understand us. At the Massachusetts Institute of Technology (MIT), scientists have designed a robot called Kismet that can have realistic conversations with people. Kismet is capable of seven different facial expressions and can vary the tone of its voice. It also adjusts its gaze and the orientation of its head towards the person it is speaking to.
Scientists at HP have designed an electronic DJ. The ‘hpDJ’ selects beats and baselines from its memory bank and mixes them. Its makers say it could be made to react to the mood of clubbers.
At the University of Texas, Dallas, researchers have designed a lifelike human face to capable of 28 facial movements, including smiling, sneering, furrowing its brow and arching its eyebrows. It could be used to put a human face to the artificial brains of the
A computer program developed at Brandeis University in Massachusetts has learnt how to design and build bridges, cranes and tables all by itself. It reinvented support structures such as the cantilever and the triangle without prior knowledge of them.
Credit card companies use a computer program called The Falcon to detect card fraud. The Falcon works by constantly updating a profile of how customers use their credit cards. It then looks for uncharacteristic patterns of credit card use in the data.
A robotic head built by a Scottish robotics company can determine a woman’s attractiveness. It works by examining faces to determine how ‘feminine’ or ‘masculine’ they are. It doesn’t work in reverse because men’s appeal is supposedly not based as much on looks. Perhaps jokingly, researchers say it could be put to use as an artificial receptionist.
Robots designed for the consumer market and employing very basic forms of AI have become increasingly popular in recent years.
Sony’s Aibo robot dog behaves like a puppy when it is first activated. But it “learns” new behaviour as it spends more time with its human owner.
A software program called FACES could stop mid-air collisions between planes. It makes planes perform avoidance manoeuvres in synch. When tested in a flight simulator, the software prevented a pile-up between 35 planes sharing airspace.
Brave new world
Over the coming century, breakthroughs in nanotechnology, the science of ultra-small machines constructed at the molecular level, may help us build more sophisticated machines that are more compact.
We may also see breakthroughs from scientists who are experimenting with connecting biological cells to silicon circuits – a phenomenon called wetware.