The rebirth of distance: Why Einstein says we should hire teachers locally

Back when I was working on internet projects at the British Council in the 1990’s, an entrepreneur, building his first business while still in Harvard, stopped by to ask us about the Japanese education market. His core assumption was that enabled by the growth of broadband in Japan, he could hire teachers in Britain, where English speakers were cheap and plentiful, and provide a better service at lower cost than schools that were stuck in Japan, paying Japanese prices.

Since then, the technology followed the trajectory he expected - even if the economics of hiring teachers in the two countries didn’t.

Faced with permanently increasing internet connectivity and ever-faster computers, it’s tempting to assume that this trend will continue indefinitely: That geographical distance will become increasingly irrelevant in determining the quality of the environment you can build for your students. But it’s also worth looking at the limitations on connectivity, and what that might mean for how we’ll have to teach in years to come.

When you transmit data over the internet, it’s limited by two things: bandwidth and latency. Bandwidth measures how much data - how many 1s and 0s - a line can carry 1s and 0s it can carry in a given period of time. Latency measures the time it takes to get a single 1 or 0 from one end to the other.

Bandwidth is often referred to as “speed”, but this is misleading; Once you’ve eliminated bottlenecks, a car travelling at its top speed won’t get down a 10-lane highway any faster than a 2-lane one. We can go on increasing bandwidth until the cows come home; Just keep on laying more cables side-by-side, and bandwidth will increase accordingly.

But latency is a lot harder. Data in a fiber optic cable currently moves at something like 2/3 of the speed of light. Unless Einstein was very much mistaken, the speed of light is the fastest it can conceivably go.

There are other factors that cause latency too - things like the time that routers take to pass data from one portion of the network to another - but even with those included, we’re already typically shifting data at between 1/4 and 1/2 as fast as God is prepared to let us go, no matter how clever we get.

You can see the effect of latency in Second Life by firing up two copies of the client side-by-side. Turn your avatar, and it moves immediately in one window, but takes a second or two to turn in the second one. This is the time taken for the information about the avatar turning to get from your computer over the internet to Linden Lab’s servers, be processed and sent back over the internet to your computer again.

Working with Second Life today, network latency isn’t necessarily the biggest consideration. There are so many ways data can get slowed down getting from one person’s screen to the other - from Linden Lab’s overloaded servers to the time it takes for our computer’s graphics cards to draw the images on our screens - that network latency is the least of our concerns. But all these things are well below their theoretical limitations. As we keep building better computers, and building out wider information highways to eliminate localized bottlenecks, the things that we can’t fix will become more important.

If you’re building a next-generation distance-learning environment to use over next 3 to 5 years, and you want to run really effective distance-learning classes, with a really responsive, as-good-as-being-there feel, you really want your teachers, and your servers (Linden Labs won’t let you do this yet) to be local.

By “local”, I mean not more than around 10 light-milliseconds, or 3000 kilometers, away from the students.