Can Microsoft Save the World?

Microsoft Research teams with top scientists to tackle the world's most pressing problems -- and it could turn conventional computing on its head in the process.

Three years ago, William Henry Gates III ordered Microsoft Research to launch a Science division. Money was one motive -- by staking out a position in the growing field of scientific computing, future profits were insured. Fortunately Microsoft Research doesn't have to turn every dollar and man-hour into marketable products. The Science group has the wonderful freedom to work on the big problems: global warming, disease, the future of medicine, the origin of the universe and the creation of life -- those sorts of things.

Leading this charge is Stephen Emmott, director of the Microsoft Research European Science Program, an Englishman with some 20 years of experience in science and computing, including a stint at Bell Labs.

Emmott's main goal is to blend computer science and traditional science, and in the process transform both. "We are at a profoundly important point in time where computer science and computing have the potential to completely revolutionize the sciences," Emmott says.

Microsoft doesn't plan to do this all alone. Today 14 Microsoft researchers are working with some 40 scientists around the world. Those numbers are rapidly expanding. "Within 12 months, there'll be 30 Microsoft Research Cambridge scientists collaborating with around 80 to 100 scientists worldwide to build new software tools for addressing important scientific challenges," Emmott explains.

These efforts were given legs during the 2020 Science conference, where some 30 scientists, hailing from nations from Japan to Germany and representing universities such as Stanford and companies like GlaxoSmithKline Inc., gathered. The group produced an 82-page oversized glossy book, "Towards 2020 Science," outlining their goals, technologies and plans.

The conference also set the stage for research projects, now ongoing, that match Microsoft researchers with their scientific counterparts.

"The real benefits come from bringing together people from Microsoft Research -- whether they're computer scientists or computational biologists or computational climatologists or oceanographers -- with people in the wider science community, to do the kinds of things that neither of us could do on our own," Emmott explains.

Understanding Life Bit by Bit
One of the more interesting Microsoft projects is "Simulating Biological Systems in the Stochastic Pi Calculus." The idea is to create a more scaleable way to track "the behavior of biological systems." One approach is to build a so-called Stochastic Pi Machine, a project driven by Andrew Phillips of Microsoft Research.

So what does stochastic mean? Well, a stochastic process is one in which there is a certain amount of randomness. This means that computers and scientists can gather data and analyze the process, but, due to its random nature, cannot make accurate predictions about it.

The Stochastic Pi Machine is designed to simulate and model the workings of biological systems. Using stochastic pi calculus, biological models can be built step by step, where models of small systems are ultimately built into a model of staggering complexity -- the type of complexity that matches the reality of biological systems themselves.


One peek at the Science 2020 work and you'd swear it was tailor made for The Bill & Melinda Gates Foundation. Not the case. While so many of the goals are identical, they are two entirely separate efforts.

"Bill has clearly had input into the overall European scientific program. He launched it last year and has provided input on numerous occasions. With The Bill & Melinda Gates Foundation there is no formal link with what we're doing. However, some of the work that we are doing clearly has an implication for the areas that the foundation is working in," says Stephen Emmott, director of the Microsoft Research European Science Program.

Nevertheless, Gates' foundation is privately funding projects that could just as easily be paid for by Microsoft Research. While Microsoft Research obsesses over the human genome, the Gates Foundation has earmarked $2.5 billion to fight malaria by building a genetic map of the disease. The foundation is also working with GlaxoSmithKline, a Microsoft Research partner, to test a new vaccine.

Beating meningitis is another item on the Gates Foundation's agenda. The foundation, along with the Serum Institute of India, is working on an inexpensive vaccine that works across all age levels and prevents the disease from being passed from person to person. --D.B.

Interestingly enough, computers are built much the same way. They start off simple, but through more and more memory, networks, grids etc., increase in complexity in an additive way.

One of the biggest 2020 goals is fighting disease, a goal shared by the Bill & Melinda Gates Foundation, which has no formal relationship with Microsoft Research Science.

"It is remarkable to think that it's 2006 and it's not known even how a cell works -- let alone how a human works. As a consequence, it's not well understood how to treat disease," Emmott says. "Once we get to a stage where we have the tools for understanding fundamental biological processes, it is only a short step to building the tools and the languages to model how disease occurs in those biological systems -- i.e. when something goes wrong or when the system is invaded by a virus."

Emmott believes that such an approach could "completely revolutionize how we think of treating disease and how we can identify and discover entirely novel therapies for treating diseases, whether they are third-world diseases such as malaria or first-world diseases such as cancer and obesity."

Projects are on the way. "We're just getting underway on a project with my team in Cambridge and one of the world's leading mathematical biologists at Imperial College in London to build a global pandemic modeling system to predict when outbreaks of diseases will occur -- global outbreaks of diseases from Avian flu to malaria. This will be a powerful tool for agencies such as the World Health Organization, for scientific researchers around the world to be able to use to do their own modeling [and] for medical research councils, as well as for interventionist types of organizations like the United Nations," Emmott says.

Climate and the environment are key factors in understanding and predicting many diseases. "There's a tremendously important relationship between environmental conditions and outbreaks of third-world diseases, but the relationship is not well understood," he says. "Being able to more effectively model and understand, say, climate change and increases in, or changes in, third-world diseases will be tremendously valuable eventually for agencies and foundations such as the Bill & Melinda Gates Foundation."

The future of science, and in particular fighting disease, lies in miniaturization. Take molecular computers. These tiny devices, while small enough to fit into a cell, are smart enough to understand inputs, such as whether a cell is diseased, and take actions -- for instance, releasing just the right amount of a cancer drug. In fact, one of the biggest benefits of molecular machines is supporting smart drugs, which can be released or held back based on their surroundings.

Sensor networks are another example of tiny yet powerful tools. Here, massive networks of sensors can be placed -- say, around a mountain -- that collect information on weather, plant life and trends, and help scientists understand the health of the mountain's ecosystem. Through wireless networks, this data can be amassed and scrutinized.


Bioinformatics: This field, also called computational biology, applies math, statistics and computer science to the understanding of biology, in particular the behavior of biological systems (see Systems Biology).
Machine Learning: Here software learns from experience. One technique is Bayesian, a machine learning approach applied to spam.
Molecular Computer: A computer small enough to fit into a cell. By detecting its surroundings and making decisions, these computers could support smart drugs.
Smart Drugs: Drugs that can adapt to their surroundings -- for instance, releasing their contents if disease is detected.
Stochastic: A Stochastic process is "fuzzy." You can collect the data and understand certain trends, but you can't predict precisely what will happen based on past experience. The stock market, especially given the last five years, is a prime example.
Systems Biology: The attempt to understand how biological components work together.

Similar to molecular computers, small, intelligent, adaptable systems could support new artificial immune systems. "Virtual human immune systems should be able to compute the results of host-pathogen interaction, including solutions to the pattern recognition problem of discriminating between self and non-self," wrote Soren Brunak, a member of the 2020 Science Group, in "Towards 2020 Science." The goal? To "compute a specific vaccine design tailored to individuals with different tissue types in the best possible way," Brunak argued.

Computational Biology, Energy & Global Warming
World health is one area where the pairing of computer science and traditional science holds tremendous promise. Future energy is another. New energy sources are important for two reasons: We are running out of fossil fuels, and these fuels, most believe, contribute to global warming. Science, in particular computational biology, could help. With this style of biology, scientists can build new tools for understanding biological processes. We could "understand how one of the world's efficient energy converters performs at such a level. There's something like 1.6 kilowatts of energy [from the sun] that falls onto every square meter of the planet every hour. The most efficient converter of that energy from the sun into its own energy for a different purpose -- for growth -- is a plant," Emmott says. "It isn't terribly well understood how they make such efficient use of the sun's energy. That's because we don't even know how a cell works, whether it's a cell in a plant or a cell in a human."

If we fully understood how plants convert energy, we could perhaps "help energy companies and energy scientists mimic that process to build entirely novel technological solutions for new sources of energy that are currently unimaginable today. That's a decade or so away but it's worth pursuing," Emmott believes.


Every year, humans crank out three times the amount of carbon dioxide that nature's "carbon sinks" absorb. If this keeps up, many scientists believe the planet will be forever changed by global warming, perhaps even doomed. While this problem is not on the top of the Microsoft agenda, Emmott's group does hope to help. "Our research efforts around climatology and earth life support systems, of which climate is an important one, are based around working with climatologists, oceanographers and ecologists, and building new computational tools that scientists urgently need," Emmott says.

"Climatologists are not short of data, so building tools to create yet more data is not urgently needed. What they do urgently need are software tools to model and couple the physical aspects of climate change, atmospheric and oceanographic type of effects, with the biotic -- produced or caused by living organisms -- elements of determining climates and climate change. That [includes] the organisms that live in the ocean that are carbon sinks, oxygen producers, regulators of the carbon cycles and regulators of climate, and also the biotic aspects such as forests -- terrestrial sources of carbon sinks and oxygen producers. It is currently largely unknown how the two interact -- the physical and biotic components. It is largely unknown what the impact is of the biotic components. It's an incredibly important component but just one that's not understood," argues Emmott.

Science Gives Back
Microsoft is also working with The Sloan Digital Sky Survey, and has already helped craft an online astronomy catalog. Physicists and astronomers for centuries have tried to unravel the mysteries of the universe -- yet after all this work our picture is far from complete.

"Understanding the universe is a large-scale data-acquisition and data-analysis problem. That is one [area] where the standard software tools that Microsoft currently produces, from Web services to database technologies to better acquire, share and analyze large scale data in the science community, can help the science community understand origins of the universe and how the universe works," Emmott says.

While science pushes the envelope of computing, computing and Microsoft also benefit.

"Science is where the real action is going to be for computer science over the next decade. By being at the cutting edge of the intersection of science and computer science, Microsoft will gain remarkable insight as to the key things the company needs to do on the broad business and personal computing challenges and opportunities a decade later. They can think of what's happening at the intersection of science and computing as being like Formula One. BMW and Ferrari do Formula One because the technology they need to develop to compete in Formula One ends up in any standard family car a decade later -- and its gives them remarkable insights into technical engineering," Emmott says.

The 2020 Roadmap is more specific. It argues that by 2015, the work done to build new scientific software frameworks will "radicalize" business computing. And beyond 2020, we should look for "novel, biologically inspired computing architectures and paradigms," according to the roadmap.

More Information


comments powered by Disqus

Subscribe on YouTube