The average three-pound human brain may lack the speed of a hard drive and microprocessor, but its flexibility outstrips that of the computer every time. Machines that can think and remember are still science fiction, but researchers are developing systems that bring them a step closer.
Computers are often used as an analogy for human memory, and there are some basic similarities. People and computers both have a working memory. Or processor, which handles information, and long-term memory, or hard drive,” where data is organized and stored. However, even the most basic home PC appears to have advantages in terms of memory capacity, reliability, and speed of access to information. Lt can store complete reference works in tiny folders and pluck out relevant details on request without the human problem of forgetting or inaccurate recall.
On the other hand, humans have a huge advantage over even the most powerful computers in our memory networking, as limitless and complex links in the brain allow us to make connections that computer designers to date can not get close to matching. Typically, computer memory, whether in an individual PC or a system such as an internet search engine – is organized similarly to a telephone directory: to retrieve a piece of information, such as a name or phrase, the computer searches through the vast amount of data to trawl up matching items. The average computer searches only for the precise item: it cannot make intelligent guesses or generalize on the basis of what it ‘ knows’ in related areas. Nor can it learn from experience as humans can – although new machines are being developed that can do this to some degree.
Expert system
Programmers can now simulate some human abilities using computer systems that incorporate the facts and rules underpinning the judgment of human experts. Such expert systems use specialized software to apply the criteria to new information and make decisions about it. Still, these can only be used in very specific areas, for example, in the diagnosis of blood diseases or the analysis of mineral samples. If the system needs more information to conclude, it checks other databases or asks the user questions. Deep Blue, one of the most powerful chess computers ever constructed, is, in effect, an expert system that applies rules and strategies gleaned from master chess players to evaluate outcomes and adapt to new strategies throughout the game.
Neural networks
At the cutting edge of artificial intelligence are computers that use circuitry that simulates the way the brain processes information, learns, and remembers. Called artificial neural networks, these computers use layers of very simple processing devices, or ‘ nodes,’ linked together to form a network that has some similarities to the networks of neurons in the human brain. The network is ‘ trained’ by being presented with examples – for instance, images of male and female faces – and being instructed on which are in which category. Over time, the network will learn to make generalizations based on its experience of the training data. It can then apply this ‘ knowledge’ when presented with new inputs – for example, when asked to judge whether a new face is male or female.
Neural networks have a real advantage in applications where there are no straightforward rules with which a more conventional computer would be programmed. For example, they are already in use in tasks as diverse as recognizing speech and analyzing financial trends. Although they are still very simple compared to the memory capabilities of the brain, these systems shiv behaviors that are strikingly similar to human memory, and they are opening new avenues for psychological research. It has been found, for example, that when the processing limits of a neural network are exceeded, it does not ‘crash’ like a conventional computer but merely performs less well – rather like a person who has too much to cope with.
A Champion Cheese Force
Even a home computer can challenge a skilled chess player, yet it took a person years of development to produce a computer that could beat a world champion. So, how did Deep Blue do it? Humans and machines play chess in very different ways. The major factor in Deep Blue’s favor was the sheer brute force of its processing power. An ordinary computer has two or three processors at most, but Deep Blue has more than 250 working in parallel, enabling it to generate up to 200 million possible moves per second. The top human players can evaluate only about three moves per second. However, as human players become more skilled, they do not consider more moves – they consider better ones. And the reason for this is memory. As they practice and study, players memorize thousands and thousands of board layouts and remember the moves that work in particular situations.
The electronic nose
Neural networks are being incorporated into ‘electronic nose’ technology, destined for use in food processing and fragrance development industries. Odors are blown over sensors that convert them into digital signals, which can then be compared and identified by the computer.
