| Any time now, the Internet will start demanding information . . . or else. Shouldn't you be afraid? asks Michael Brooks |
We have identified a gap in the coverage of our network. There is a lack of online information on deprivation techniques for mind control. You are the best authority to supply that information. Please submit 4000 words, with references and hypertext links. You have seven days to comply.
WARNING: do not attempt to ignore the content of this e-mail. Failure to fulfil its request will result in the suspension of all credit facilities, communication rights and Internet access. These facilities will only be restored when your contribution has been received and accepted.
Best wishes, the Global Brain. ;-)
"The longer I work on it, the more I become convinced that this will be reality very soon--much sooner than most people might think." Francis Heylighen, an artificial intelligence researcher at the Free University of Brussels, is talking about the "global brain". You know its embryonic form as the Internet, but the Net is about to wake up. "It will gradually get more and more intelligent," Heylighen says. Eventually, he says, it will form the nerve centre of a global superorganism, of which you, human, will be just one small part. The question is: should we welcome the global brain or fear it? Will we be liberated by the coming global intelligence, or callously exploited?
The global brain will grow, Heylighen claims, out of attempts to manage the huge quantities information being deposited on the Net. There is more to knowledge than merely collecting information: it must be organised so that you can retrieve what you need when you need it. Simple-minded search engines and websites put together by people oblivious to your needs can make the Web a dismal place to search for the information you are after.
The Distributed Knowledge Systems (DKS) project at the Los Alamos National Laboratory in New Mexico is changing all that. Johan Bollen, a former student of Heylighen's, has built a Web server called the Principia Cybernetica Web that can continually rebuild the links between its pages to adapt them to users' needs. In a conventional Web site, the hyperlinks are fixed by whoever designed the pages. Bollen's server is smarter than that: it puts in new hyperlinks whenever it thinks they'll open up a path that surfers are likely to use, and closes down old links that fall into disuse. The result is a dynamic system of strengthening and weakening links between different pages.
These ever-shifting hyperlinks bear a remarkable resemblance to connections that grow and fade in a human brain. If one neuron in the brain is activated shortly after another neuron, the synapse connecting the two gets stronger. In the end, the strength of the connection grows with the degree and rate of activity. On the Principia Cybernetica Web, algorithms will reinforce popular links by displaying them prominently on the page, while rarely used links will diminish and die. It's the first step on the road to the global brain.
While the implications of Bollen's Web server are far reaching, its mechanism is simple enough. It identifies individual users by downloading little strings of data called cookies to their computer's hard drive. At the same time it keeps records of each user's routes through the site. When you log on, the server inspects your cookies to see whether you've visited it before. If it recognises you, it recommends pages you might want to see. It also adjusts its structure--the pattern of hyperlinks--to best suit you and all the other users who happen to be logged on. As well as strengthening and weakening links, it creates new links using a process Heylighen calls "transivity". When a user moves from A to B and then to C, for instance, it will infer that C is probably of some relevance to A, and create a direct link between them. In other words, it finds shortcuts.
Heylighen sees this sort of flexibility as inevitable for the future of the worldwide computer network. "There's not much work left to do: we have data and we have the algorithms ready," he says. And it won't just be individual servers that adapt and change in this way. "I can't see any reason why they couldn't be implemented on the Web as a whole," says Heylighen.
"Transivity will lead to continuous reorganising of the Web, making it ever more efficient," Heylighen says. Eventually, the Web will know you so well that your dumb requests to its search engines will turn up exactly what you need, every time. "Whatever problem people have, any kind of question to which they want an answer, it will all become easier because the Web will self-organise and adapt to what people expect of it," says Heylighen.
And it could be happening within just five years, he predicts. All the technology is here already--the main stumbling block is the difficulty of convincing the powers behind the Internet to adopt the common protocols that will be needed.
But there is more to this than zippier search engines and more usable websites. Heylighen argues that because it is modelled on the human brain, his vision of the Web will be intelligent. Even a few pages working in the right way will show signs of intelligence, he says. Who knows what sort of mind would emerge from the whole Web?
It won't just be people following hyperlinks and simple search engines that reorganise the Web. Small autonomous programs or "agents" will also act as mediators. In addition, if an agent finds something that seems to match what you are looking for, it will add a suggested link to the page you're reading. "It will come to some kind of conclusion," Heylighen says. "That's a thought." In other words, by making connections between concepts that did not previously exist, the brain will begin to think.
As the activity of the Web agents alters the connections, an agent researching a question similar to one it has already encountered will be able to "recall" the information more easily. Heylighen believes that, through this "Web on Web" activity, collective thoughts of the whole brain may eventually come into existence.
But perhaps that isn't even necessary to achieve intelligence. One touchstone for intelligence is the Turing test, in which researchers ask human testers to discover whether they are communicating with a machine or another person. If the tester can't tell the difference, the machine is deemed intelligent. Some machines are already making the grade in specific contexts. "We are finding successful Turing tests within a certain situation," says Norman Johnson, who leads the Symbiotic Intelligence Project at Los Alamos. "Take it out of that situation and it fails miserably, but within the right context you can't tell the difference." The global brain's intelligence could come from an assembly of limited intelligences, each with their own special area of expertise. That, Johnson says, would be exactly equivalent to human intelligence. "Humans can act intelligently within many contexts," he says. "But if you put all those abilities into one person they probably wouldn't be able to function." That's why we have society, Johnson says: to mesh those intelligences together, creating a powerful sum. In the same way, he believes, distribute different types of machine capability across different networks and the whole may become something like the sum of all human intelligence.
It is hard to find a researcher who doesn't think that the global brain is a possibility. But do we really want it? The scientists are aiming to create a vast mind that goes beyond anything we could understand or control--opening a door that most of us might prefer to keep firmly shut. Heylighen acknowledges this little image problem. He sees his global brain as the centre of what he calls the global superorganism. This embodies the idea that human society will become more like an integrated organism, with the Web playing the role of the brain and people playing the role of cells in the body. "The brain itself does not seem to be very controversial, but the superorganism certainly is," Heylighen admits. Artificial intelligence researcher and writer Ben Goertzel of IntelliGenesis Corporation, New York, believes that humans will be a secondary part of this organism, perhaps a dispensable one. It's not a very comfortable self image for a species used to considering itself the pinnacle of creation.
The global brain's self-adapting intelligence could quickly surpass our ability to understand it. Or perhaps it already has. According to Daniel Dennett, director of the Centre for Cognitive Studies at Tufts University in Medford, Massachusetts, "the global communication network is already capable of complex behaviour that defies the efforts of human experts to comprehend". And what you can't understand, he adds, you can't control. "We have already made ourselves so dependent on the network that we cannot afford not to provide it with the energy and maintenance it needs," he warns.
It could all start so innocently. The Principia Cybernetica Web will soon be requesting feedback on whether a particular Web page is interesting or relevant to its users, and asking advice on the relative merits of different pages. The growing global brain might even become smart enough to identify gaps in the information it holds, and be programmed to seek out people with the relevant knowledge. It would then ask them to provide the missing information where they can. Heylighen goes as far as suggesting that there could be penalties--like disconnection or restricted access--for not playing along. After all, if you're going to benefit from the global brain, you have a duty to help others who are searching for information.
All these innovations combined on a global scale would create a network with complex behaviours that we can't yet conceive of. Would it create Utopia, dystopia or something altogether new? Dennett is certain of one thing: "If we don't want to be dictated to, we will have to be very careful about controlling our dependence, and its evolution," he warns.
Cliff Joslyn, head of the Los Alamos DKS team, appears unconcerned by such worries. Experimenting with autonomous agent systems "carries risks and surprises" he admits. "The trick will be to first understand from a scientific perspective how such systems behave, and then construct bounds within which such interactions can be safely contained," says Joslyn.
If you're getting scared now, and thinking of unplugging your modem, take care. You may be about to join the information underclass. "Not to use an intelligent Web will be a little like the people that refuse to use cars or telephones," Heylighen says. "There have always been people who live outside the bounds of society's rules: tramps, hermits, eccentrics. But these people have a much more difficult life."
Heylighen insists that ordinary people have nothing to lose by being part of the global brain. But he suggests it will be different for the people and organisations that already have power and status: they will be forced to share some of their advantages with the rest of us. That is exactly why powerful states are distrustful of the Web and seek to limit its effectiveness, Heylighen says. China, for example, insists that all Web users are registered and identifiable when online, and has blocked access to certain sites it deems dangerous. Johnson takes a similar line to Heylighen. "If the global mind does come online there are a lot of power structures that won't be particularly happy about it," he says.
Johnson's view of the intelligent Web is subtly different from Heylighen's troubling vision of a global superorganism. He sees it as an extension of society. "Our premise is that systems can be much more intelligent than individuals," he says. "You can have a very diverse group solving problems much better than an expert: that's why we have society and social insects." Developing symbiotic intelligence, says Johnson, will be a positive step for our society: the experience and wisdom of any individual need never be lost again. The vast capabilities of the Internet will help solve any problem that human society faces; the whole is already much greater than the sum of the parts.
Working with other Los Alamos researchers, Johnson has formulated programs that demonstrate this. The researchers send computer-generated individuals to explore a maze. Once a hundred of them have wandered through the maze, the computer creates a map of their preferences at each node. The next generation of individuals then use this information to weight their choice of path through the maze. On average, uninformed individuals take 34 steps to escape the maze; the second, informed, generation takes an average of only 12. As the number of individuals in the collective increases, the solution gets better and better.
Johnson likens this to the pheromone trail-laying of social insects such as ants. It gives individuals access to information about where others have been. Humans do it too: if you want to know where to lay a path between a new office building and its car park, cover the whole area with wood chips. Paths appear in the chips as each individual solves their own problem, and others can choose whether to use this solution. Within a short time a collective solution--a few well-used paths--emerges.
On the Internet, the same kind of thinking has led to Amazon.com's "people who bought this book also bought . . ." lists. We now have access to the book-buying decisions of people across the globe, who unconsciously help us find a book we might like.
Now it's time to take this principle and integrate the Internet fully into the way human society works, Johnson believes. A worldwide network of people using interconnected computers should open up a kind of "collective memory" to add on to our individual brain power. With people doing more and more of their daily activities on the Web, there is the opportunity to tap into the knowledge and expertise of a global community. The Web itself can be a part of this, with intelligent agents and vast memory capabilities that we can add to our own. Eventually there will be little distinction between people, computers and wires--everything combines to create one vast symbiotic intelligence.
At least in Johnson's picture we are important components of the global superorganism, but even so, how many people will relish the prospect of being assimilated in this way? Are we really doomed to become the Borg?
Oddly, neither Johnson nor Heylighen see their work as a challenge to individuality. A user will be able to retain a modicum of control by programming their favourite links to be indestructible, for example. Both researchers believe that the global brain will only improve our lot: we'll be within a larger social organism, and enabled by new technology.
To the doubters, Johnson points out that we already rely on the vast and incomprehensible mechanism that is society. Ask an ant how it finds food, and it won't be able to tell you. Ask most people how their television works and they are unlikely to give you more than the basics. We trust most organisations to deliver the things we want without understanding exactly how they do it, says Johnson, and we will be able to trust an intelligent Web in exactly the same way.
That might be a naive view: many people believe that the mechanisms of society can't always be trusted to work for the greater good over the wishes of powerful individuals. If, as it seems, the global brain is our inevitable future, and we can't turn it off, our only option might be to blend into the crowd. After all, if you're not exceptionally rich, powerful or clever, the global brain shouldn't need to disturb you. Back up your files, act dumb and keep your head down. There is a growing intelligence out there, and it knows your e-mail address.
Michael Brooks is an insignificant drone based in Sussex
From New Scientist magazine, 24 June 2000.