Microsoft brains go head-to-head with Google

Microsoft has hired top scientists in a quest for search algorithms that will allow it to compete directly with Google

Microsoft is actively working on new search algorithms it will use to power its own search engine to take it into competition with Google, according to the head of the company's Theory Group.

Speaking in Sydney at the Fifth International Congress on Industrial and Applied Mathematics, professor Jennifer Tour Chayes said Microsoft is patenting new search algorithms with a view to replacing the Inktomi technology currently powering MSN's search with Microsoft's own.

"Since Yahoo acquired Inktomi, Bill [Gates] has decided we need our own capacity," she said, adding that the company is already patenting new algorithms it believes have the potential to power a new search engine.

Microsoft last month launched a new search program called MSNBot, which trawls the Web to build an index of HTML links and documents -- functions previously left to Inktomi and other partners.

It is believed to be the first step in a multi-year plan to build new search technology that encompasses home and business users, with the ultimate goal of the technology being to bind Microsoft's various Web sites, applications and the Windows operating system.

Beyond search, Chayes believes filtering is the next killer application that will require the input of high mathematics. "As computers become more pervasive, we are going to be assaulted as we walk around," she says. "To take advantage of all the good things IT has to offer, you will need a filter. We can't all have a secretary do the filtering for us, so there'll be a theoretical solution."

Founded to perform blue sky research -- one of Chayes' speeches to the ICIAM is entitled "Phase transitions in combinatorial optimisation" -- Microsoft's theory division has already contributed directly to product development, notably the new version of Active Directory shipped with Windows Server 2003.

"The product team asked us for assistance with a bottleneck where an algorithm was preventing us from working with larger networks," she said. "One of our team, Laszlo Lovasz, developed an improved algorithm in a day."

Despite this incredible service, the theory group is not a resource available on tap within Microsoft. "We have people dedicated to interfacing between the product and research groups, so that when a problem comes to us it is very well described" Chayes says.

Perhaps the best reason for this is the stature of some of the team's researchers: Lovasz is a winner of the Wolf Medal, one of the world's leading mathematics prizes, and is also a former Yale Professor. Others in the eight-strong permanent team have won the Field Medal, mathematics equivalent of the Nobel Prise.

And what of phase transitions in combinatorial optimism? Chayes explains that understanding how phase transitions -- for example from liquid to gas -- can be expressed mathematically may make it easier to solve problems like load balancing servers, so that limited processing resources can be applied to multiple tasks more efficiently.

Understanding this phenomenon, she says, means a thorough understanding of how water boils may help enterprises wring more out of their servers.

"We're a very far out group at Microsoft," she says. "It could be 50 years from now before some of this work is used."

Jim Hu and Mike Ricciuti contributed to this report