Microsoft science fair hits Silicon Valley

Updated May 7, 2:12 p.m., to correct the spelling of Hal Schectman.

MOUNTAIN VIEW, Calif.--Microsoft packed up TechFest and took it on the road Thursday, bringing a chunk of the company research unit's annual science fair to Silicon Valley.

TechFest, and the smaller TechFair event that is taking place here this week, are designed to allow product teams at Microsoft to better understand what folks in the research labs are up to--and vice versa.

Just how important are Microsoft's research labs? Well, the head of the unit notes that Microsoft started the research effort in the early days of personal computing.

"Microsoft is still here," Rashid said. "Virtually all of Microsoft's peer companies from that era are gone or not recognizable. Ultimately it is about survival."

The event here featured some hits from this year's TechFest--including a "translating telephone" as well as some projects of local interest, such as an effort to provide more data to scientists studying, among other things, methods to protect fish in San Francisco Bay Area waterways.

That latter project started out more than three years ago as an effort to monitor conditions along the Russian River, two hours' north of San Francisco, and has expanded to cover all of California. The idea was to use computers to mash up a number of different data sources from weather to temperature to rainfall to sediment patterns, and present the results to researchers in an Excel spreadsheet.

The challenge isn't a lack of data, said Microsoft researcher Catharine van Ingen, noting that sensors, satellites, and cheap computational resources have added up to "an unprecedented ability to get your hands on data."

Now, she said, the challenge is being able to take all that data and mash it together in ways that are useful.

"It's not just finding a needle in a haystack," she said. These days, she said, it takes work just to find the right haystack. Microsoft's tools, for example, are making possible a deeper analysis of the many factors that go into deciding what makes for ideal conditions for migratory salmon.

"We can start asking questions we couldn't ask before," said James Hunt, a University of California at Berkeley, researcher who has been using Microsoft's tools.

The same underlying system is also being used to monitor seismic data in the Swiss Alps and weather data in Southeast Asia, noted Rick Rashid, the head of Microsoft's research labs.

Rashid said many of the research projects eventually find themselves into shipped products, such as the instant answers and many of the other features that are part of Bing.

In addition to the press, there were customers and even a few rivals in the crowd inside the Galileo auditorium at Microsoft's campus. Among the notable attendees was Alan Eustace, Google's senior vice president of engineering and research.

Next door to the auditorium, roughly two dozen projects were on display in the demo room. One of those perusing the booths on Monday morning was Chuck Thacker, who earlier this year won the Turing award, the highest honor in computer science. Thacker said life has been good since getting the award, but with the recognition has come a lot more requests for him to speak. And while he will be doing more of that, he has had to turn down some opportunities, such as a request that he appear at a conference in Turkey on econometrics. "I know nothing about econometrics."

Thacker also said he has had less time to work on his current research efforts.

"It's a drag," he joked. "Fortunately, I have a lot of good colleagues."

Rashid chimed in, "It's a nice burden to have."

Among those presenting their work at TechFair was Cynthia Dwork, whose work focuses on how to make sure that when researchers offer aggregate data that individual privacy is maintained. Historically, those collecting data have aimed to remove truly personal information like names, phone numbers and Social Security numbers, and assumed that was enough.

However, that's not real privacy, Dwork notes. Quite often, those who already know something about a person in a group could identify that person's record using their existing information.

"It's so easy for these systems to accidentally leak information when an attacker knows something about you," Rashid said.

Truly private data, according to Dwork, is that in which the data being shared is nearly the same even if one person is removed from the data set. Dwork presented a mathematical formula to demonstrate how to tell if a particular data set is provably private or not. "The definition holds no matter what the attacker knows," she said. "We don't care what other people know, we don't care what their computational capabilities are."

Other projects included one that is exploring a mobile scenario in which a smartphone could elect on the fly whether to run parts of a program on the phone itself or send the computational work to a remote server or PC, taking into account factors such as the phone's processing power, battery level, and the available network conditions. Yet another was exploring whether one might be able to create usable panoramic still pictures from a rather grainy video clip.

Some technology Microsoft showed off at TechFair were things that had already graduated from the research labs into shipping products, such as Bing's social search features that allow the search engine to display near real-time results from Twitter within the search product.

Showing that was Hal Schectman, one of the many search people who has joined Microsoft from Yahoo in recent weeks as part of the two companies' partnership in that area. Schectman showed how Twitter could augment a search result for "Iron Man 2" by bringing in real-time sentiment. The results being displayed were all from the last 15 minutes, though Schectman said that's not good enough.

"The goal is to have this down to seconds instead of minutes," he said. "Actually, one second would be our goal."