One of the most exciting – and troubled – computing projects in recent years has been One Laptop Per Child. It aimed to produce a $100 laptop designed specifically for children in developing countries, with a production run envisaged to run into many millions. For various reasons – not least political ones to do with its use of GNU/Linux rather than Windows – that hasn't happened, but the resulting OLPC XO machine has, nonetheless, proved an important stimulus to work in this area thanks to its radical design.
On the software front, that included rethinking what a GNU/Linux distribution should offer, and how children with little or no experience of technology could use free software to learn about computers and everything else. Its hardware was equally innovative, notably in terms of its wireless mesh network, and revolutionary screen.
The latter was designed by the OLPC's founding CTO, Mary Lou Jepsen, whose background is in optics and screen technology. Here she talks about how the XO design arose, her new startup, Pixel Qi, which aims to build on her work for the OLPC project to make a $75 laptop possible, and why the most important thing about open source in this context is not that it is free, but lean.
GM: What were the technological advances that allowed the XO to be created for such a low price?
MLJ: Really, the most important thing was to go back to a conceptual tabula rasa – a clean slate – and to not try to make the OLPC XO machine happen by cost-reducing a current-generation laptop. These are not, as it were, Dells on a Diet.
We started by thinking about the display – because that is how the typical user experiences – literally, sees – the computer and it’s the most expensive element of the laptop. I designed the XO screen to be high resolution – because school involves lots of reading! I decided that the screen, unlike those in conventional laptops, had to be sunlight readable – because many developing-world students study out in the open. And it had to be help conserve power – because that was critical. In very rough terms half of homes in the world don't have electricity. Focusing then on power conservation led to eliminating the hard drive, in favour of Flash memory. No moving parts!
And then we focused on turning off the computer’s microprocessor when the user stops running an application. The screen has to stay legible – so that the user, the student can keep on reading, studying, a book or a website, even though the computer has stopped. Then, when the user starts typing again, the computer has to resume operations as quickly as possible. We got the suspend time and the resume time down to the point where the user can't notice it, in 1/10th of a second.
These are just examples – and there are many. A top executive at one of the world’s tech giants told us we were trying for too much: you’re asking for 20 miracles – and we have a rule around here: one miracle per product. But we were not doing a product, we were pursuing a cause, and it turns out that 20 miracles is about right for a cause.
GM: How much of this was Moore's Law, or knock-on effects of it?
MLJ: Not much – if anything, we turned in a different direction. Moore’s Law is the idea that the number of logic junctions you can build into a chip increases, rapidly, over time. This Law describes the great gains in CPU complexity - that fueled an enormous increase in the power of all sorts of electronic machines. However, it has also allowed software to become inefficient. Indeed, the XO architecture’s critical considerations for the CPU, the microprocessor brains of the machine, include… how fast can you turn it off? This is because we aimed for a very low power architecture – and that means turning the CPU off as quickly as the XO senses that the CPU isn’t needed. This is a bit like the Toyota Prius. One key innovation in the Prius is that it shuts off the petrol-burning internal combustion engine as soon as the car stops. That’s really similar to what we’re doing in the XO. Adding power to the CPU in this analogy is rather like adding more cylinders to the engine in the Prius: you could do that, but it doesn’t make much sense.
Also, the power of CPUs grew so quickly that a CPU that was state-of-the-art a couple of years ago runs decent applications – and still has plenty of value for most potential users. There’s an interesting idea in business called the Innovator’s Dilemma. A typical company incrementally improves its products – until they no longer meet the needs of an emerging class of customers. That’s really what’s going on here. Conventional laptops got more and more powerful – and are no longer relevant to many classes of needs.