There are over a billion school-aged children in the world. Nicholas Negroponte wants every one of them to own a laptop, whether they’re in Maine or Mali. To reach this goal, Negroponte plans to build 100 million laptops in 2008, doubling the global production of laptops. His engineers are designing a new type of machine, a laptop designed for the needs of children and the challenges of computing in environments without reliable power or network infrastructure.
Negroponte’s One Laptop Per Child project is far from the only effort dedicated to bringing personal computers to the developing world – a recent World Bank report found more than twenty low cost computing initiatives underway worldwide. Not all initiatives are charitable – companies like AMD, Intel and Microsoft are warming to the idea that there’s money to be made in selling computers in developing nations. But while there’s great excitement about low cost computers for developing nations, the failure of previous attempts suggests that producing an inexpensive computer that people want to use is harder than it might appear.
Does bringing low cost computing to the developing world require the invention of new technology? Or new business models to make existing technology available in poor countries? Is this a technical or economic challenge? Will governments and philanthropists bring computers to the developing world, or will market forces?
Phones or PCs?
To understand why engineers, philanthropists and marketers see developing nations as the next PC marketplace, it’s worth examining the worldwide success of mobile phones. The Gartner Group predicts the sale of 850 million new handsets in 2006. 200 million of those phones will be purchased in China and India. Phones are selling in poorer nations as well – over 100 million people in sub-Saharan Africa own a new or used mobile phone.
The impact of mobile phones in the developing world has been sudden and profound. Farmers use phones to check market prices and schedule deliveries. Businesspeople replace travel to meetings with phone calls. A study from the London Business School suggests that small increases in the teledensity of nations is correlated with increases in economic success – because wireless phone networks are easier to build than wired ones, entrepreneurs are blanketing the developing world with cell towers and selling millions of handsets.
In comparison to mobile phones, the global PC market is quite small – 172 million machines sold in 2004, most in developed nations. It’s less clear that people are clamboring for PCs in the developing world the way they’ve sought out mobile phones. For many people, the economic impact of a PC is more distant than that of a phone – a phone means crops can sell for a higher price tomorrow, while a PC means a child may have a chance at a better job in the future.
But some things are easier to do on a PC than on a phone. Holding a Java-enabled GSM phone in his hand, Walter Bender, President of the One Laptop Per Child nonprofit, asks, “Written any code on one of these things?” The answer is obvious – while new applications are developed every day for mobile phones, the applications are programmed on personal computers. For an educational project which plans to teach the world’s children computer programming and replace textbooks with a handheld device, the mobile phone looks like an imperfect solution. But for companies hoping to capture consumer spending in developing nations, competition from increasingly powerful mobile phones may be a fearsome barrier to PC adoption.
Changing the Economics of the PC
Microsoft believes that PCs are less popular in the developing world in part for economic reasons. While a hundred dollar handset is within the purchasing capacity of many Brazilian households, a thousand dollar PC is not. Microsoft has introduced a new system – FlexGo – which the company describes as “pay as you go computing”, intended to make traditional PCs more affordable in middle-income nations. Rather than paying a thousand dollars for a PC, a buyer pays a small fraction of the price up front, then buys prepaid cards that allow her to use her computer. When time expires, the computer locks itself, using hardware designed by Microsoft’s hardware group, until the user purchases additional time via a prepaid card. Once the user has purchased 800 hours worth of prepaid cards, the computer unlocks and works as a conventional PC.
Based on a successful pilot of the FlexGo in Brazil, Microsoft is now launching additional pilots in other middle income nations (China, Hungary, India, Mexico, Russia, Slovenia and Vietnam). Possible partners include Internet Service Providers, who might offer the devices for free as part of a subscription service – the cost of the PC would be included in monthly Internet service payments, and the ISP could disable the computer if the owner fell behind on payments.
It’s unclear whether many users will be willing to trust their data to a device that can lock its owner out. While the payment plan lowers the upfront cost of the machine, the FlexGo PC ends up costing more than a conventional PC due to finance charges. Other than the novel payment enforcement mechanism, the FlexGo PCs don’t appear to have any other adaptations for developing world conditions. This is likely by design – Bill Gates and other Microsoft executives have been dismissive in commenting on lower-cost computing devices, suggesting that people in developing nations should share conventional computers rather than using lower powered alternatives.
AMD has taken a radically different approach to selling computers in middle income nations. Targeting the same market as FlexGo – middle-income nations, marketed through internet service providers like Cable and Wireless – AMD’s (now discontinued) Personal Internet Communicator looks radically different from a traditional PC. Shaped like a grey brick (customized by different ISPs with colorful plastic accents), the computer is designed for more rugged use than a traditional PC and optimized to use little power, featuring AMD’s Geode firstname.lastname@example.orgW Geode processor, which draws roughly one watt of power. The low power consumption allows the machine to operate without a fan, a critical design consideration, as cooling fans are prone to failure, particularly in dusty environments, like Saharan nations.
The specifications for the PIC likely won’t have too many existing computer users trading in their desktop machines: a 366Mhz processor, 128MB RAM, 10GB hard drive, 56k modem. The PIC doesn’t come with native support for wireless internet access or ethernet – ISPs are selling the machine with an adaptor that turns one of the devices’ four USB ports into an ethernet jack. AMD compensates for the lightweight hardware by running a customized edition of Windows CE, a version of Windows designed for handheld devices. Unlike most instances of Windows, Windows CE is offered as “shared source” software, allowing manufacturers like AMD the ability to partially customize the environment. The environment used on the PIC includes an “XP Extender”, allowing users to use some Windows XP applications.
Cable and Wireless, the primary Internet service provider in the Caribbean, offers the PIC with mouse, keyboard and CRT display for $350, discounted to $300 for high-speed internet access subscribers. C&W also provides financing options, allowing customers to purchase the device for 24 monthly payments of $15. A Turkish ISP offers two packages, one with a CRT and another with an LCD panel monitor, for $280 and $440 respectively – both come with a foldable Razor-style scooter thrown into the deal for good measure!
The choice of monitors highlights one of the shortcomings of the PIC for use in the developing world – while the PIC itself uses very little power and can be run off a 12 volt car battery as well as AC power, external monitors draw lots of power. CRT monitors draw roughly 150 watts when in use – LCD monitors are more efficient, drawing roughly 50 watts… but both dwarf the power requirements of the PIC itself and mean the PIC isn’t an appropriate solution for users without a regular electricity supply.
AMD recognizes that the PIC won’t be the solution for the entire developing world – they’re selling the product as the “first step” in their wider 50×15 initiative, designed to provide connectivity to 50% of the world’s population by 2015. Having a product on the market in developing nations gives AMD the opportunity to test their product and pricing models and position themselves as a leader in the low-cost computing field. And providing PICs for charitable efforts is winning them converts. Professor Gary Chapman from the LBJ School of Public Affairs at the University of Texas set up a lab of PICs for Katrina refugees and reports, “We plugged 50 of them into a single power cord, which just isn’t something you can do with a conventional PC.”
One challenge the PIC will likely face is the falling price of conventional PCs and the low price of recycled PCs. Chinese firm YellowSheepRiver is selling a significantly fuller-featured PC (DVD drive, wifi, larger memory and hard drive) for roughly $150, using Linux rather than Windows CE. Recycled PCs, scrapped in the US or Europe but revived to run Windows 98 or Linux, are often available for under $50. The real question for the PIC may be whether it’s sufficiently different from existing options to develop a market in developing nations.
If you build it, will they come?
The FlexGo and 50×15 efforts suggest that selling computers in emerging nations may largely be a matter of economics, pricing machines at an affordable level and providing financing options. But adjustments to pricing won’t be sufficient to provide computers for the very poor, for people without regular access to electrical power or for users who don’t speak languages supported by Windows or Linux. To bring computing to rural India, the founders of the Simputer Trust decided to break away from traditional PC models, creating a new, hand-held computer designed specifically for rural Indian environments.
The Simputer was conceived at the Global Village seminar, part of a 1998 IT conference in Bangalore. The core team of engineers who designed the product included several prominent computer scientists from the Indian Institute of Science at Bangalore. They took on an extremely ambitious set of goals: the machine had to be useful for users who were illiterate, had to have low power requirements, needed to be under $200 per unit, and each unit must be shareable by a group of farmers or merchants to defray costs.
The design that emerged was a handheld computer, similar in size to HP’s iPaq. To support illiterate users, the Simputer would be capable of speech synthesis in languages like Hindi and Kanada. Users could customize the device with a smartcard that would store personal data and settings, letting each user of a shared device have their own personal desktop. Rather than build a company to produce the Simputer, the designers settled on an unusual structure for the effort – the hardware and software design are owned by a trust, and licensed to manufacturers who want to build the machine.
The idea of the Simputer – open software and hardware, an indigenously designed and produced technology for India, a computer for people not thought of as computer users – was widely celebrated in Indian and international media. But the actual Simputer device was slow to reach the marketplace. The Simputer founders had a difficult time attracting investment and manufacturing interest and weren’t producing large batches of commercially available devices until 2004, three years after much of the hype about the project. A major selling point of the device – the ability to run on AAA batteries – was abandoned due to poor battery life, and the machine was redesigned to run on rechargeable lithium ion batteries. This made the device less useful for environments off the electric grid, the original audience for the device.
By 2005, only four thousand devices had been sold by two manufacturers, PicoPeta and Encore Technologies, well below the anticipated sales levels. PicoPeta has relaunched the device as the Amida Simputer with a color screen and support for connections to CDMA mobile phones – paired with a mobile phone, it has features competitive to smartphones like the Palm Treo and adaptations to support Indian languages. But it’s unclear that the Simputer will be able to compete against the falling prices of smartphones, and increasingly clear that the device is less a computer for rural communities than another option for mobile, urban professionals.
The difficulties the Simputer Trust has had in bringing products to market offers at least two possible lessons for projects that seek to bridge the digital divide with novel hardware: the global press certainly wants a low-cost computer for the developing world, but individuals may not be willing to purchase one.
In developing a low-cost educational laptop for the One Laptop Per Child project, Nicholas Negroponte – co-founder of MIT’s Media Lab – has not been publicity-shy, developing momentum for the project through public demonstrations of the device for global leaders like Kofi Annan. His sales strategy avoids the problems of consumer choice by selling the devices a million at a time to national ministries of education for deployment throughout national school systems. As a result, the “pilot” phase of the project promises to roll out a million laptops each of four different developing nations – Brazil, Libya, Argentina and Nigeria.
While the One Laptop Per Child project is sometimes referred to as the “hundred dollar laptop” project, the ambitions of Negroponte and his team go far beyond that of producing an inexpensive computer. Walter Bender explains, “It’s a laptop for children, as opposed to being a cheap laptop.” Specifically, it’s a laptop for children who may not have access to regular electric power or telephone lines, which requires the laptop to have features not found in devices costing thousands of dollars.
To allow the machine to be manufactured by Quanta, a leading Taiwanese laptop manufacturer, at a price of about $135 per unit, governments need to invest in a million laptops at a time, leading some wags to observe that it’s a “$135 million laptop”, not a $100 laptop. For the investment to be attractive to the Education ministries the project is approaching, the device is being marketed in terms of textbook replacement. Rather than purchasing paper textbooks, books will be made available on the laptop, and the money spent on textbooks over four or five years will be used to purchase the device.
But for the device to meet the goals of the designers, it needs to be a device children want to use, experiment with and discover. So the machine is designed with “game console” and “home theatre” modes as well as eBook and laptop modes – the laptop hinge allows the screen to be folded flat to the laptop body for use as a book or game console. Because the motherboard of the machine is mounted directly behind the screen, the hinge contains only the wires to the keyboard, not the large number of fragile connections embedded within a traditional laptop hinge.
Since students will be dependent on the laptop for multiple years as their primary textbook, it needs to be extremely robust. The machine shell is being designed of 2mm thick plastic, as opposed to the 1.3mm used for most commercial machines. When folded, it seals with a rubber gasket, and the rubberized keyboard is highly resistant to dust and liquid. With no hard drive, cooling fan or fluorescent monitor backlight, the components that most frequently fail in laptops are absent from the design. And the case is designed to be assembled and disassembled quickly, both to increase ease of repair and reduce cost of assembly.
Many of the decisions made by the OLPC team are designed to make the laptop as power efficient as possible. While the machine can be run from traditional AC or DC power sources, it’s been designed so that a child generate sufficient power to operate the device. Humans aren’t very efficient generators of electrical power – an average adult, riding a bicycle, may be able to generate 20 to 30 watts over a sustained period of time. A small child, using her arms to power a hand crank, might be able to produce 5 to 10 watts over a very short period of time. This presents a major problem for powering a contemporary laptop, which can consume as much as 30 watts of power, depending on what tasks it is performing – ten minutes of hand-cranking might produce three minutes of power for the laptop.
The design goal for prototype laptop is a machine that could be used for 10 minutes for every one minute spent generating power. This requires designing a laptop with very low power consumption needs – the target for the first generation of the laptop is 2.5 watts while the processor and color screen are active. Using a low-power processor – the same AMD Geode GX email@example.comW used in the PIC – CPU power draw is radically reduced. When the computer is being used as an eBook, the display buffer stores a copy of the screen being displayed – this allows the CPU to shut down until a new image needs to be produced. Displaying in greyscale, with the processor shut off, the machine should draw half a watt of power.
Many children will be able to charge their laptops at their schools, either through connections to the electric grid, or via pedal-power generators that children could take turns riding. But giving each child a laptop means giving every child the means to power a laptop whether or not she has access to electricity. Early designs of the laptop included a handcrank, inspired by the design of the Freeplay human-charged radio. But handcranks use small muscle groups in the arm and proved difficult for children to use for sustained periods of time.
Squid Labs, an engineering firm co-founded by Media Lab alums Saul Griffin and Colin Bulthap, is working on prototype power systems. The current design features a microgenerator powered by a pullstring, like the one uses to start a lawnmower. The user holds the generator – about the size of two hockey pucks stacked together – in her left hand, grips a handle in her right and pulls a meter-long cord. The cord spins a fine shaft at roughly two thousand revolutions per minute, generating power in the laptop’s batteries. As the person pulling the cord tires, an embedded microcontroller adjusts the flow of power to the battery so that the generator operates at maximum efficiency while the generator speed slows.
Adults and children over twelve are able to produce 20 watts using the generator for short periods of time, but can produce 10 watts for almost unlimited periods of time. Children under twelve tend to tire quickly when generating power. Bulthap predicts an economy of power generation developing around the laptop and generators: “Whether it’s an older brother, or a strong guy in the village, small kids are likely to turn to an older person to power their laptops. This may turn into a business for some people as human powered devices become more common.”
This power generation strategy requires intelligent battery choice. Lithium ion batteries, favored for use in high power laptops for their efficiency and power density, don’t tolerate voltage spikes well. Human-generated power is necessarily spiky – people generate more power when pulling a string than when resting between pulls – so the laptop uses nickel metal hydride batteries instead, which tolerate the spikes, are easier to dispose of in an environmentally safe fashion, and are cheaper as well.
Griffin notes that there’s more uncertainty about the ability to manufacture power systems than there is about any other part of the laptop: “Companies build 50 million laptops a year, but over the span of history, we’ve only built five million human powered electrical appliances.” But Bulthap believes the device can be manufactured for $10 apiece using Chinese factories to produce the device. Eventually, he’d like to see the technology produced in countries where it’s used, and generators available to power a variety of devices: laptops, mobile phones, lighting systems, televisions.
Aside from the processor, the most power-hungry subsystem of a laptop is the display. A conventional LCD screen is also expensive, costing $130 in the average laptop – with a total price target of $135 for a laptop, a conventional LCD isn’t an option for the OLPC project. Former CTO of Intel’s display division, now CTO of the One Laptop project, Mary Lou Jepsen has looked at a number of unconventional strategies for the laptop display, including eInk, a low-power, high resolution display technology used in Sony’s LIBRIé e-book reader.
But eInk displays have only been produced in small quantities. By contrast, she notes that the “thin film transistor LCD manufacturing base is larger than the fab base for RAM memory chips or contract silicon manufacturing.” A solution that’s inexpensive and can be manufactured swiftly involves using the existing LCD manufacturing base, but revisiting “the LCD sandwich”, the combination of color filters, liquid crystal and transistors that power a conventional laptop display.
Jepsen singles out the color filters as the main culprit in contemporary displays – they absorb 80% of the light created by a fluorescent backlight, wasting power, and represent a third of the manufacturing cost. To create color without filters, Jepsen and a team at Taiwanese diplay manufacturer Chi Mei are experimenting with colored LEDs as light sources, creating colored pixels without filters. Other technologies include lenticular displays, plastic sheets etched with microscopic lenses (produced much the same way a DVD or CD is printed) that focus white light separated by diffraction gratings, and techniques that use less expensive partial color filters.
Jepsen ran a “beauty contest” of competing display technologies in the fall of 2006, and chose a display that allows the team to meet the specifications: a 7.5 inch diagonal screen capable of displaying 1200×900 pixels in black and white mode, and SVGA-quality (800×600 pixels) in color. To be useful in outdoor classrooms, where students might meet in the shade of a tree, the black and white display needs to be readable in bright light – Jepsen reports that prototypes the team is producing have 30% reflectance, rivaling e-paper technologies. Because the OLPC group is patenting much of the work on the monitor, it’s been difficult to get specifics on the exact technology in use – Jepsen says her current plans for the device focus on “a combination of diffractive and field sequential solutions in an LCD without polarizers,” which would work quite differently from existing laptop screens but be capable of being manufactured in existing LCD fabrication plants.
The largest technical challenges to the laptop’s success may not be power or display, but networking. Rather than deploying laptops to classrooms along with spools of ethernet cable, the laptops are equipped with 802.11s-compatible wireless radios, allowing a set of laptops to establish a mesh network. This lets students in the same classroom share a virtual whiteboard with a teacher, chat during class or collaborate on assignments. If the school has a connection to the Internet via phone or satellite, a server/router can be connected to the net connection and share connectivity with other nodes in the mesh, acting as a peer to the laptop nodes. Equipped with a hard disk, the server/router could cache documents, an important capability, as the laptops have no disc drives, and use 512MB of flash RAM to store 130MB of applications and operating system plus any documents.
“A thousand node mesh will have maybe a few kilobits per second for each user, but that’s vastly better than no connection at all,” notes Michail Bletsas, OLPC’s chief connectivity officer. Michail is designing the laptops to connect not just in school, but when students bring them home as well, blanketing villages and parts of towns with mesh networks. The laptop’s “bunny ears” are external wifi antennas that provide a 3db gain. Mounting them vertically, rather than horizontally as they are in most laptops, gives an additional 2db gain. “Consumer laptop manufacturers don’t have to worry much about signal to noise ratios. They’re counting on getting a wifi signal through scattering. But we’re dealing with a very different situation in a village environment: fewer hard walls, fewer ceilings.”
Bletsas believes his design will provide node to node connectivity over 600 meters. To participate in the mesh, a student doesn’t need to open her laptop – she just needs to flip the antenna up. This turns on the wifi subsystem of the machine without waking the CPU, allowing the laptop to route packets while consuming about half a watt of power. Students are likely to learn a great deal about mesh networks very quickly – if you fail to turn up your antennas or keep your laptop powered up, two of your friends won’t be able to chat through your node later in the evening. And they’ll be learning on a network that’s using IPv6, allowing Bletsas to avoid the complications of network address translation.
Like the Simputer, the OLPC project is pioneering software as well as hardware. RedHat Linux has been modifying Fedora to meet the needs of the project, a set of needs unfamiliar to many Unix developers. The project “goals” page reminds participants that “this is a laptop that is designed for children, not for system administrators or programmers” – this means that “children must be allowed to play”, and any mistakes they make that cause a machine to become unstable have to be easily reversible, perhaps through something as simple as a reset button.
The optimization also requires some deep changes in how Linux usually operates. Because flash memory has a finite number of read/write cycles before it decays, standard filesystems which rewrite their data structures in place repeatedly are problematic – Fedora, implemented on the OLPC hardware, will use the JFFS2 filesystem, which supports “wear leveling”, a technique for spreading wear to the memory throughout the full 512MB of the storage chip.
“The fundamental design goals for the software of the project are to give students and teachers tools that leverage their ability to learn, their ability to be expressive and their ability to be social,” Walter Bender explains. The laptop will include drawing and music software to allow students to create their own media. The desktop environment – codenamed “Sugar” – tries to break down the isolation students might experience from staring at laptops all day by introducing the idea of “presence”. The desktop is aware of other students in the classroom, showing their pictures or icons on the screen, allowing students to chat or share work with others in the class. New tasks – a drawing, a document – are shared by default, though students can choose to make them private. Sugar creates a “blog” for each child – a record of the activities they engaged in during the day, with the ability to add public or private diary entries.
Beyond the Device
How will teachers react to the introduction of a tool that allows students to talk to each other during class, annotate text and surf the Internet? Will teachers embrace the participatory education models advocated by OLPC advisors Seymour Papert and Alan Kay? Or will the laptops be viewed as a threat, with students forbidden to use them for anything other than reading ebooks?
Bender acknowledges that the ambitions of the OLPC project go well beyond creating physical devices and require an overhaul of educational systems around the globe. “There’s a reason the machine is a laptop,” he notes – even in school systems which don’t take advantage of the full capabilities of the machine, children will be able to take the systems home and learn with the machines.
Intel’s Eduwise initiative, which appears to have many of the same goals as the OLPC project, is planning to address this problem by reaching teachers first, and has announced a project to provide 300,000 Mexican teachers with laptops, in cooperation with the Mexican government. Development on the Eduwise device – called “Classmate” – is in a much earlier stage than that on the OLPC laptop, but Intel’s Paul Otellini has promised a “full-featured” laptop, capable of running conventional Windows software, available at a price point of $400. The strategy of equipping teachers first might help teachers learn to use PCs as teaching tools – the OLPC initiative is pushing a model where entire schools, teachers and students, get computers at the same time.
Questions that go beyond the hardware and software of low-cost computing devices will likely go unanswered until pilot programs bring large numbers of computers into developing world schools. The OLPC project plans to deploy 5 to 10 million laptops in 2007 in five nations, which will surely reveal some of the “ecosystem” problems the introduction of low-cost devices into developing nations may cause.
Computing in the developed world involves not just the millions of devices in use in homes, schools and offices – the ecosystem that supports computing includes internet service providers, software stores, repair technicians, and recycling facilities. As projects bring low cost computing to the developing world – whether through government purchasing or private enterprise – the challenge of creating this environment will fall at least in part on these technology pioneers.
While hardware engineers are creating novel displays and power generation systems to allow low cost computers in the developing world, similarly innovative discoveries will be required to ensure that these new devices can be distributed, maintained, repaired and responsibly disposed of at the end of their lifetime.
Joris Komen, director of Schoolnet Namibia, an award-winning educational computing project in southern Africa, asks in an open letter, “Will Nicholas’s laptops have the cost for ISO-14000+ compliant end-of-life recycling factored into their cost-price? This is an important consideration, given the fact that there are no ISO-14000+ compliant recycling plants anywhere in Africa.” Komen goes on to note that his project spends $5 per machine they dispose of to ship the machine back to Europe so it can be correctly disposed of – to support disposal of the millions of laptops OLPC plans to distribute, the project team may need to work with governments to create recycling centers.
The issue of computer repair may be even more pressing. If a child must have a laptop to participate in lessons, it’s critical that a broken or stolen laptop doesn’t mean the end of a child’s education. Will businesses emerge to repair low-cost educational computers, or will schools be expected to repair the machines that break in the course of use? The colorful case of the OLPC device is designed not just to be appealing to children – it signals that an orange or green computer being used in a developing world business is a machine that’s been taken from a child, using social pressure to deter theft. But in countries where a $100 machine represents a substantial fraction of annual income, it’s easy to imagine that some of these machines will make their way from children’s hands into the marketplace.
Worries about problems like computer disposal, repair and theft signal acceptance of a radical idea: the idea that millions of people in the developing world will purchasing personal computers in the next decade. Increasingly, the questions about computing in the developing world no longer begin with “if”, but with “when”.
Can the PC spread as rapidly as the mobile phone? Will consumers buy PCs, or will children discover the devices first in schools? Will the developing world run on Windows and conventional hardware, or on Linux and low power hardware?
If all goes as planned, in a few years you can ask any schoolchild in the global South – they’ll turn on their laptops, connect to the Internet and let you know.