Sharing what matters
Jun 7th 2007
From The Economist print edition
Software: A computing maverick hopes to upgrade the web, transforming it from a document collection into a data commons
MOST people find it difficult to keep up with Danny Hillis’s imaginative leaps. In the 1980s he dreamt of building intelligent computers and co-founded Thinking Machines, a firm with a mission to make machines “that will be proud of us”, as he used to put it—with tongue only half in cheek. That did not quite happen, but Mr Hillis did, in the process, pioneer the field of massively parallel supercomputing. After a stint at Disney, where he proposed building a theme-park full of free-roaming robot dinosaurs, he turned his attention to building a mechanical clock that will run for 10,000 years, a task that arguably requires genius in its justification as well as its execution. Now this maverick of the technology industry has a new idea that could have a big impact rather sooner than that.
It concerns the web, a creation that, though impressive, is pedestrian compared with what Mr Hillis has in mind. Today’s web allows easy and universal sharing of documents. Before the web, internet users could share documents only by making bilateral arrangements—requesting a document from someone else by e-mail, for example—which incurred transactional “friction”, so that relatively few people did so. The web eliminated that friction. Today it is obvious that this was world-changing, but Mr Hillis still remembers “how hard it was to explain” before it happened.
Déjà vu. The next step, he says, is to let the web do for data what it has already done for documents. Just as there used to be lots of people with interesting but unshared documents, today there are innumerable people and organisations with useful but locked-up databases. These range from topics of life-and-death importance—the World Health Organisation’s data on bird-flu outbreaks, say—to things that are deceptively banal but potentially useful—a foodie’s private spreadsheet listing the best wines at his local restaurants, say. For data to change the world as documents have changed it, the web must again eliminate all friction involved in sharing.
Metaweb Technologies, a firm set up by Mr Hillis and his co-founders, Robert Cook and John Giannandrea, aims to do exactly that with Freebase, a website that sits on top of a new kind of database. The name is not a pun on cocaine but a contraction of “free” and “database”, since the database shares the spirit of Wikipedia, the free and collaborative encyclopedia. (Mr Hillis is on the advisory board of Wikipedia’s parent organisation.) Just as Wikipedia lets people contribute information to its articles, Freebase, which is in a test phase, will let anybody contribute, correct or recombine data. The difference is that information on Wikipedia tends to be “unstructured”—ie, buried in text—whereas on Freebase it will be structured, so that each item can be re-used in any context.
It is an open question whether enough people will contribute their data to generate the momentum of Wikipedia, but Mr Hillis is optimistic. “Most people with data want others to have and use it,” he reckons. A boffin who collects data on butterflies, say, might want to upload it so that others with the same fascination can add their own information. Another researcher might then add data on lizards, and yet others might then combine the data on butterflies and lizards with existing geographical data to create maps or analyse patterns. The fact that users will not know in advance how their data might be used is precisely the point.
This requires a new level of flexibility in the database. When building most databases today, programmers decide in advance what sort of questions users might wish to ask of the data, by defining what are known as the “schema”—the types of records in the database and the relationships between them. Metaweb’s 35 programmers, by contrast, have built a new sort of database, based on a more flexible structure known to programmers as a “graph”, which allows users to contribute and use not just data, but schemas as well. They can, in short, ask any sort of question of the database.
Metaweb is thus very different from commercial database software, such as that made by Oracle, and from Google Base, which might superficially appear similar because it too allows anybody to upload data. Google Base, says Mr Cook, consists of many independent data sets that are stored in a coherent way. This means that many records are duplicates—if several people upload the details of the same digital camera, say—and may even contradict one another. Metaweb, by contrast, reconciles conflicting data and ensures that each object exists only once in the database. But each object can be tied to every other object, so that the resulting web of associations looks rather like the neural networks in a brain.
There is one similarity to Google, however. The search giant’s founders, Sergey Brin and Larry Page, initially concentrated on perfecting a technology (search) that could change the world, without worrying about a business model, which came much later (in the form of advertising). Mr Hillis plans to do the same. For now, he is much too excited about the technology to worry about the money. “Everything else I’ve worked on, if it succeeds, only helps one thing,” he says. “This has the potential to make everything better.”