Sunday, May 11, 2008

Return from Microsoft

The past weekend I attended a special event organized by the Microsoft Live Search team. The Live Search team invited 28 up-to-graduate PhDs from US and Canada to come to an on-site interview. The uniqueness of this event is that nearly none of the 28 candidates have been telephone interviewed by the team before and many candidates (including me) even have not submitted a resume to Microsoft before being invited. The Live Search team searched out these candidates (probably through the Live Search) and invited us. The theme over the event is straight and clear---beat Google!

About the Trip

In general, the trip is full of pleasant. The Microsoft recruiting team have organized a wonderful event. It is warm, joyful, and with exciting surprises. Thank you, Erin Bucholz (my direct contact recruiter), Jared Singer, Rondell Honcoop, Ben Mercer, and others (sorry I am not good at memorizing names).

Each of the 28 candidates is a selected one with a particular background that is related to Web search. For instance, there are candidates with the background of query optimization, distributed computing, image processing, data mining, natural language processing, and so on. My particular background is labeled Semantic Web. Moreover, I found that I was the only candidate invited to this event with the primary background as Semantic Web. It is exciting to be in a group of young scientists of varied disciplinary areas while at the same time with a focused general theme.

Microsoft has arranged most of the candidates (including me) staying at Westin Bellevue, a modern 5-star hotel that is adjacent to upscale shopping centers. The general environment of the hotel is superior. The guestrooms are luxuriously furnished in modern style---clean, simple, straight, and with modern decor.

At the very morning of the interview day, Microsoft sent a limousine (a 14-passenger Navigator model I believe) to pick up the candidates to the campus. It was a slight surprise to everybody and we were joking to attend a party rather than an interview event.

limousine exterior
limousine interior

The Interview Event

In this interview event, the Live Search team has scheduled three group discussions and four-round individual interview for every candidate. All the three group-discussion leaders are well-known Web researchers from Microsoft Research. They are Yi-Min Wang, Chris Burges, and Paul Viola. There are totally more than 10 individual interviewers. Each of them is a program leader with a particular focus on Live Search. Microsoft was indeed serious to this event.

Yi-Min Wang is a passionate speaker. He has a strong passion on competing Google. At the beginning of his session, Wang briefly introduced himself and described a general paradigm of Web search from the industrial point of view. In short, Web search is about finding a matching between billions of Web pages and billions of search queries, while at the same time the numbers of both pages and queries are increasing.

After the brief start, the rest of the session was focused on questioning and answering. In particular, when answering a question Wang described his experiences on fighting to fake Web pages, which was once reported by The New York Time. This topic also led to many discussions of a broader issue of the online advertisement business model and the risk of this popular business model.

I asked a question to Wang how he thought of the factor of humans in his described grand picture of Web search. His answer was primarily focused on the user interface issue. Nevertheless did I agree with his points, he had neglected mentioning the connections between Web resources (such as data, services, and links) and the humans who create or own these resources. I thought that the factor of these connections should at least be another critical issue with respect to my question. But time for question answering was limited and thus he might just not have enough time to expand his discussion.

The session with Burges was slightly different from the previous one. Burges began by asking every candidate why they came to this event. I answered by quoting myself a motto---"We may beat Google, by not by following the Google way." Microsoft is thinking of defeating Google on Web search. Hence I am very interested in coming to hear how Microsoft would approach this goal and I am willing to share with Microsoft how I think this goal could be approached. I would be glad to join them towards this goal together.

The addressing of "beating Google" does not mean at all, however, that Google is bad or evil. It is only about that we need competition to improve Web search better and better. Eventually Web users will be the biggest winners. With the same purpose, I have shared my viewpoints with the CEOs of Mechanical Zoo and Imindi (two ambitious startup companies but with great potential) in the last two weeks. Max Ventilla (Mechanical Zoo), Adam Lindemann (Imindi), and I have shared this common belief---by not following the Google way, we may approach alternative fascinating solutions for Web search and Web knowledge reorganization. Again, I shared this motto with Burges and he told me that it was also exactly what he believed.

Burges did a brief slide show for us about his new role at the Web search team and the Microsoft Live Search ranking algorithm. After his talk, I asked how he would compare the Microsoft ranking to the famous PageRank algorithm used by Google. He replied that the real ranking algorithm used by Google has already not been the original PageRank for long time. The very core of the Google ranking algorithm is a top secret. But Microsoft is catching up.

I agree with Burges. The PageRank algorithm is too raw for real-world products. To get high quality search results, Google must have done a significant revision of this general algorithm. The revision might have been so great that the eventual Google ranking used now may actually no longer be named the PageRank in its standard sense.

On the other hand, however, the PageRank algorithm reflects the Google's philosophical vision of World Wide Web. Google evaluates the Web to be a network of linked nodes where the importance of individual nodes is primarily determined by the linking popularity inside the entire network. This philosophy is the fundamental of "the Google way." Hence by "not following the Google way" it means that we need to think of the Web in a fairly different picture from the one Google thinks. I have such a different picture described. Adam Lindemann of Imindi shares a very similar vision as mine. But what is the picture that Microsoft thinks? Burges had not directly addressed an answer, and neither had Wang. Unfortunately, I have not gotten another chance to explore this issue deeper with another Microsoft developer or researcher in this trip.

In the third session, Paul Viola did a formal presentation to all the candidates in lunch. His talk was primarily on how to do a good research particularly in an industrial company rather than in an academic institute. He made several constructive suggestions for young scientists and engineers. The talk is informing and helpful. Due to the time constraint, however, we have no chances to ask him specific questions.

After the lunch, we came to the most fun part of the event---the swimsuit competition (i.e., individual interviews).

All the interviewers are excellent leading developers and researchers at Live Search. They are very well experienced and more important, they all have the passion on what they are doing (at least for the four who had interviewed me).

The four-round individual interview is scheduled into four main topics---research background discussion, coding, algorithm, and design issues. At each round, a candidate got to talk to an interviewer for about 45 minutes. In general, I have done a fair but not an excellent interview. I have described my thoughts to them, written a few lines of code, and solved some problems. At the same time, however, there was a problem I could not figure out the final answer to the end. I had tried to think of it from all angles except of the angle the interviewer looked for (and it is not a familiar territory of mine, what a pity).

In short, I had emphasized two points about the future of Web search: (1) the switch from the "God" role to the "Guru" role of search engines, and (2) the importance of "proactiveness" in the next-generation Web search. I emphasized to the last interviewer (a kind lady) that we were not just searching the Web. By contrast, we are searching the evolving Web; and this is a key when thinking of beating Google. If Live Search can think of the Web evolution a step further beyond Google, Live Search would get a better chance to beat Google.

Final Address

The final decision of this interview event will come to me in few days. No matter whatever, however, I appreciate this oppotunity and it gives me a chance to hear the first-hand opinions about the future of Web search from the most frontier industrial developers and researchers as well as from many peer PhD candidates all over the North America, let it alone that I had a wonderful journey at Seattle.


Anonymous said...

And it seems like they take all means to beat Google in search: :-)

Yihong Ding said...


Thank you for referring me the link.

Yeah, Microsoft is really planning to competing with Google at the realm of Web search. Eventually, the market Microsoft wants to get in is online advertisement. Microsoft believes that the dominating of Web search is the key to take the largest slide of the huge online advertisement cake.

Certainly this thought has no problem. But the strategy Microsoft has taken is questionable. Buying either Yahoo or just Yahoo search can bring immediate short-term benefits to Microsoft. But in the long term, you will see how this purchase may eventually hurt the company.

In fact, Google should not worry about the deal between Microsoft and Yahoo. The reason is somehow interesting.

If this Microsoft-Yahoo deal happens, the future of Microsoft search become predictable and Google can certainly be still ahead with its superior research and development power. However, if this deal could not happen, Microsoft must try its best to seek the real weakness of Google and develop real novel approach on Web search to defeat Google. Then the future becomes unpredictable. Based on the power of Microsoft research, Google should be really worried.

Microsoft-Yahoo deal will actually great restrict the future of Microsoft search. This is why Google should actually "encourage" it to happen.



gregory said...

don't they know that the define themselves by what they oppose? rather than defining themselves but who they are?

you want to beat google? why?, is the first question. ego? coexistence works too.

why not want to make something unique and good? that at least has a chance to get somewhere

funny big old company. and here is a prediction, in not too many years, google will be in the same space, trying to find someone to imitate.

here is a suggestion, spin off your top twenty guys, give them a hundred million and their own chalet ... only requirement, they cnanot call the home office for a year