4 ms·
The deep learning is that there are forces that lead to centralization: generally they are variations of human choice and seeking economies of scale. For insta
by michaelfeathers 10y ago
The deep learning is that there are forces that lead to centralization: generally they are variations of human choice and seeking economies of scale.
For instance, it's no accident that we started with hundreds of thousands of small websites and ended with most traffic going to a few. Ask yourself how the internet could have been constructed that would have prevented sites like Facebook or Twitter winning the popularity contest. The 'Power Law' dynamics behind this are the same ones that lead to some airports being hubs and others not, the fact that there is a backbone rather than everything going point-to-point, and many other phenomena.
We can have decentralization but there are costs. Someone has to pay them.
- pessimizer 10y ago> The deep learning is that there are forces that lead to centralization: generally they are variations of human choice and seeking economies of scale. The reason that some airports are hubs and some are not is due to the the fact that it's a system spanning large areas of varying densities. Distance doesn't exist much on the internet, other than firewalls. The goal of most internet companies is to create that distance through lack of interop, in order to guide the maximum amount of traffic through a single point where it can charge a toll. It's typical rent-seeking. I'd argue that a bureaucratic metaphor is far more apt than any spatial one.
- michaelfeathers 10y agoI think it's more graph theory than distance. A fully connected graph costs more than an minimal spanning tree.