Pushing the limits

Andrew Herbert’s job is to think the unthinkable. As Microsoft UK’s research chief, he and his 100 staff are at the cutting edge of technology. Jon Excell reports.

Once a year, at its headquarters in Redmond, Washington, IT giant Microsoft hosts an event that must seem like nirvana for computer geeks.

At this gathering — known as Techfest — its army of international researchers show the rest of the company what they’ve been up to, arm their colleagues with ideas and inspiration and generally soak up the plaudits for a life spent at the cutting edge of innovation.

This year’s event, held earlier in March, did not disappoint — and was notable for being the first time Microsoft invited the press, providing a fascinating glimpse into the future of the company, and to a certain extent, the future of computing.

Microsoft was the first software company to create its own research organisation, which has developed a character apparently at odds with its parent company’s corporate philosophy — its researchers’ academic candour contrasts with the tight-lipped corporate approach found elsewhere in the organisation.

Nowhere is this balance between the academic model and the business of product development more seamless than at the company’s oldest research outpost, here in the UK at Cambridge University. And the relationship between academia and industry is embodied by Andrew Herbert, the group’s managing director.

After receiving a PhD in computer science from Cambridge, Herbert spent a number of years in the university’s computer science department before striking out into industry and eventually joining Microsoft in 2001. Speaking to The Engineer from Techfest, he outlined the core areas of research carried out by his team of more than 100 researchers.

A large part of the group’s focus is on what might be considered Microsoft’s bread-and-butter areas. For example, the system networking group is concerned with developments in operating systems, networks and distributed computing. A key area here, said Herbert, is developments in multi-core processors, a growing trend where multiple processors are attached to integrated circuits to improve computer performance.

Another group that reflects the company’s traditional concerns focuses on ideas that underpin the design of programming languages. ‘We have a strong interest in techniques to enable us to produce software that’s been mathematically verified to do exactly what it says on the tin,’ said Herbert.

The centre also hosts a group that is carrying out research into machine learning and perception. It is looking at the design of algorithms for applications in areas such as computer vision, image recognition, information retrieval and handwriting recognition.

Herbert explained that the approach to these problems has changed in recent years. While computer scientists once attempted to program computers to work like the human brain, the emphasis today is on modelling the physics of the world using statistics, then building robust statistical algorithms.

He said: ‘We’ve built systems that will look at images and say “that’s a picture that has a car in it, that’s a picture that has a house in it”, but the computer isn’t being intelligent.

‘It has been shown many pictures of cars and built up a statistical pattern. It’s saying “this is like all the other things you showed me and if you said those are cars then this must be a car”.’

Herbert believes the development of this approach could be key to the future of computing.

‘One of the limits to computers is the screen and keyboard way of interacting,’ he said. ‘If we want to have computing around us doing more things, we’re going to need to be able to speak to computers and they’re going to need to be able to recognise gestures and to accept handwriting as input.’

Straying further from Microsoft’s traditional area is the computer mediated living team, which is studying the role that computing technologies can play in everyday life. This group is involved in the ethnography and sociology of how people use technology and is, for example, interested in the many ways mobile phones can be linked into technology in the home.

‘They have a strong interest in the kind of things that might appear in your sitting room or kitchen that make interacting with modern technology easier,’ said Herbert.

The work of this group is particularly well suited to the aspirational nature of events such as Techfest, and work from the Cambridge lab was very much in evidence in Redmond.

One of the technologies on show was Bubbleboard, an attempt to turn the impersonal answering machine into something more social and interactive. ‘It’s basically a screen,’ explained Herbert. ‘if a message is taken, a bubble appears on the screen, old messages sink to the bottom, while new messages float to the top. You can use your finger to move the bubbles around and group them together — we’re exploring things like that.’

The group also unveiled digiPostcard, a screen divided into little blocks to which members of the family can send pictures from mobile phones. The aim of this demonstration, said Herbert, was to illustrate how technology can be used to help people stay in touch in a less intense and more sociable way than emails and telephones.

Herbert admitted that such systems are designed to provoke discussion and could be many years away from being turned into production items. However, elsewhere in the group, researchers are investigating technologies that could reap more immediate dividends for the company.

In the most recent initiative, Herbert has set up a computational science group that is examining the increasingly important role computers are playing in everything from exploring the origins of the universe to understanding genetic systems. With scientists becoming more dependent on computers to simulate experiments that would be impossible to carry out in the lab, this group is working closely with the European scientific community to develop a range of new tools.

Herbert said the work of the group is likely to be particularly relevant to Microsoft’s burgeoning healthcare business. While this area has hitherto been concerned with products relating to healthcare administration, he said that computational science could ultimately have a huge impact on the clinical side of things.

He described how, in an early success, the group has taken a tool used for modelling software and adapted it to model biological systems at the cellular level. While the technique has already been used to build a simulation of some of the basic components of the human immune system, Herbert speculated that scaled-up versions of such systems could ultimately revolutionise healthcare. ‘Your computer could be running a model of you, and your doctor, when thinking about how to treat conditions, would effectively design a treatment for you,’ he said.

Looking beyond his own group, a popular theme that caught Herbert’s attention at Techfest, and which perhaps hints at a future assault on Google’s dominance, was the amount of work going into personalising internet search systems.

‘If a search system knew more about your personal interests and understood the task that you were doing at the time, then it will get more accurate information for you,’ said Herbert.

‘The World Wide Web has given us a complete new dimension to computing but it is a bit of a wild west out there. How we get some structure in the world is very interesting.’