• Choose a category
• All
• Classic-Fiction

By Root 425 0

or thousand-year revolution. We need to imagine a world where the construction of the scientific information commons has come to fruition. This is a world where all scientific knowledge has been made available online, and is expressed in a way that can be understood by computers. Imagine, furthermore, that the data aren’t isolated in tiny little islands of knowledge, as they are today, with separate, siloed descriptions of phenomena that are fundamentally connected in nature, phenomena such as amino acids, genes, proteins, drugs, and human medical records. Instead, we’ll have a linked web of data that connects all parts of knowledge. Rather than mining that knowledge in a piecemeal way, we’ll be able to do automated inference on all of human knowledge, finding hidden connections on a scale that dwarfs the work of Swanson or even the SDSS. We’ll give this dream a name: we’ll call it the dream of the data web.

The data web sounds grandiose. But, as we’ve seen, we’ve already taken many small steps toward the data web, through projects such as the SDSS and the Human Genome Project. What’s gradually emerging is an online netcalof knowledge that’s intended to be read by machines, not by humans. Those machines will find meaning in that network of knowledge, and help explain it to us. In the remainder of this chapter we’ll ask how the data web will be built, and what it will mean.

There is, however, a difficulty in the discussion, a difficulty that bedevils every discussion of the potential of computers to find meaning in knowledge: the more you speculate on this potential, the further you go in the direction of a discussion of full-blown artificial intelligence, the science-fictional the-internet-wakes-up-to-take-over-the-world type scenario. That’s a lot of fun to talk about, but it’s too easy to get bogged down in speculative questions: “So, can machines ever become conscious, and what is consciousness, anyway?” or “Well, yes, maybe one day the internet will wake up and take over, and what of it?” This is all ground that’s been trodden many times before. Instead of repeating those discussions, we’ll explore a middle ground between the near-term projects discussed earlier in the chapter, and full-blown artificial intelligence. This middle-ground future is conceptually rich, fascinating, and strangely under-discussed, perhaps because the dreams of artificial intelligence exert such a strong pull on the imaginations of the technologically curious. What we’ll do is synthesize current ideas from computer science to understand what happens when you take today’s algorithms and imagine a future in which they can be applied across all scientific knowledge. As we’ll see, the likely results are spectacular.

Data-Driven Intelligence

To understand what the data web can be used to do, it helps to give a name to the ability of computers to extract meaning from data. I will call that ability data-driven intelligence. Examples of data-driven intelligence include the algorithms used in the Medline searches Don Swanson did to discover the migraine-magnesium connection, the algorithms used to correlate Google searches with CDC flu data, and the algorithms used to mine the SDSS for dwarf galaxies and orbiting black holes, and to discover the Sloan Great Wall.

The term “data-driven intelligence” is not new. But at present it is mostly used in a more restricted sense than what I’m proposing, to describe data-driven approaches to making corporate business decisions—for instance, the way airlines mine data on passenger no-shows to know how much to overbook their flights. I’m proposing to use the term in a much more general way, as a broad category of intelligence, similar to the way we use terms such as “human intelligence” and “artificial intelligence.” In this general sense, “data-driven intelligence” is a much-needed term, partly because of the large and rapidly growing number of examples of data-driven intelligence. But what’s even more important is that the term highlights a particular approach to finding meaning, an approach for which