The view from Silicon Valley

As I start this column, I face a blank computer document in OpenOffice, the Linux equivalent of Microsoft Office. As I type and save, the hard drive of the machine I'm using (a server belonging to a service called Workspot) begins to accumulate information. When I'm done, my home directory at Workspot will contain more information than when I started. If you charted my progress at intervals, you'd see a rising curve as the information accumulated. You might choose to describe the curve as a gradient: a procession from a time where there's no information, to one where there are some 600 words putting over an idea – that is, information. On the web, there are gradients over time, but also from address to address. But there's a bit of a twist.

Consider Google: relative to, say,, or almost any other internet address, Google has a huge amount of information, and therefore a very large information gradient relative to other sites. But in another sense, Google itself doesn't have much of a gradient. The information there is just as easily accessed by one user as another: there are no "secret" enclaves of information accessible only to people who type in the right request. (As far as we know.) Gradients in information space seem to depend both on the amount of information, and its accessibility.

In physical space, a given amount of substance can't be moved without changing a gradient. Not so in information space: the same information can exist in two places at once – say, on my home page and in Google's cache. It is normal for information to increase, and to spread, and thus for information gradients to be relatively flat in open systems such as the web. But steep information gradients are possible, too. Imagine a genius weblogger working in isolation: if no one links to her site, the information is unavailable and the gradient is steep. The first reader to land on her site would have an advantage over others in the use of our genius's work. But once many other sites link to her, the gradient goes away. Everyone can get to the information, and the advantage is widely available. This is how weblogs work: find one that interests you, and you'll quickly find the best and brightest 'blogging in that field.

Put another way, shallow or flat gradients are democratic: everybody has equal access to the benefits that the information may confer. Steep gradients on the other hand, are advantageous to the few at the expense of the many.

Steep gradients might be bad or good, depending on which side you find yourself. (Think of people inside and outside brokerages selling shares during the boom.) Over time, however, steep gradients are, by definition, bad for the majority, and good for a minority. The rise in prosperity on our planet has tracked the spread of democracy, leading me to believe that shallow gradients are good for all in the long run, and steep ones are bad.

So the web is good, weblogs are good, and Google, as long as it behaves the same for everyone, is good. But projects such as the eavesdropping Echelon and the US's "Total Information Awareness" project are bad, since they represent steep gradients. It doesn't matter how well-intended such projects are: if the gradient is steep, over time they benefit a few at the expense of many.