Microsoft's Bing Mantra: We are looking for fundamental change in structure of the Web
Microsoft CEO Steve Balmer has given Qi Lu plenty of resources. Lu's job is to develop Microsoft's search engine Bing as a serious competitor to Google.
By ET Bureau | Updated:
Qi Lu had grown up in China too poor to afford decent meals regularly. His fortunes changed when he attended a talk by Carnegie Mellon professor Edmund Clark, who was impressed by Lu’s questions.
Clark persuaded him to apply to Carnegie Mellon University, and it eventually led to a PhD and then jobs at IBM and Yahoo. Lu led Internet search at Yahoo for 10 years . He quit in 2008 to start a company or return to China, but was grabbed by Microsoft to lead its search efforts.
Microsoft CEO Steve Balmer gave Lu plenty of resources and several years of support. Lu’s job is to develop Microsoft’s search engine Bing as a serious competitor to Google.
This may not be easy, but Lu is eminently qualified to make an attempt, as his association with search is as old as the technology itself. Excerpts from an interview with Lu by Hari Pulakkat.
Why is search so hard?
You can think of search in many ways as rocket science for the masses. The web is so vast, there are trillions of URLs and people consume the web through very simple interfaces, typing a short query. You need to understand the user purpose based on a very short form of input.
What inputs can you have, apart from the keywords?
Today, most of the information we use is based on keywords. Location is increasingly becoming important, particularly for mobile searches. We have this model called the W4 model – who, when, where, what – that is a strong predictor of possible purposes.
The time (of a query) turns out to be a strong predictor of information need. For mobile devices you have a clear indication of who the person is. With increasing social network connections, we can look at ‘who you’re with’. And now ‘what’ is also another key source.
For example, on mobile devices you may be using a calendar, and we have an additional source of information . We aspire to build technologies for the next few decades so that we can systematically bring knowledge at the time when people need it.
There are two technological pillars that represent this quest. First, the computational understanding of user intent. How do we use software and computational models to understand the purpose? Our second pillar is called knowledge. The web is evolving into a full-blown digital society. It will substantially outgrow the intellectual heritage of its reservoir documents.
We want to understand the purpose of our users, understand the digital universe and use those to empower every human being with knowledge to enrich their lives.
Clark persuaded him to apply to Carnegie Mellon University, and it eventually led to a PhD and then jobs at IBM and Yahoo. Lu led Internet search at Yahoo for 10 years . He quit in 2008 to start a company or return to China, but was grabbed by Microsoft to lead its search efforts.
Microsoft CEO Steve Balmer gave Lu plenty of resources and several years of support. Lu’s job is to develop Microsoft’s search engine Bing as a serious competitor to Google.
This may not be easy, but Lu is eminently qualified to make an attempt, as his association with search is as old as the technology itself. Excerpts from an interview with Lu by Hari Pulakkat.
Why is search so hard?
You can think of search in many ways as rocket science for the masses. The web is so vast, there are trillions of URLs and people consume the web through very simple interfaces, typing a short query. You need to understand the user purpose based on a very short form of input.
ADVERTISEMENT
What inputs can you have, apart from the keywords?
Today, most of the information we use is based on keywords. Location is increasingly becoming important, particularly for mobile searches. We have this model called the W4 model – who, when, where, what – that is a strong predictor of possible purposes.
The time (of a query) turns out to be a strong predictor of information need. For mobile devices you have a clear indication of who the person is. With increasing social network connections, we can look at ‘who you’re with’. And now ‘what’ is also another key source.
For example, on mobile devices you may be using a calendar, and we have an additional source of information . We aspire to build technologies for the next few decades so that we can systematically bring knowledge at the time when people need it.
ADVERTISEMENT
There are two technological pillars that represent this quest. First, the computational understanding of user intent. How do we use software and computational models to understand the purpose? Our second pillar is called knowledge. The web is evolving into a full-blown digital society. It will substantially outgrow the intellectual heritage of its reservoir documents.
We want to understand the purpose of our users, understand the digital universe and use those to empower every human being with knowledge to enrich their lives.
ADVERTISEMENT
There’s a lot of data on the computer of the person who is searching. Is it possible to look into that?
Our belief is a clear ‘Yes. But it will take many, many years of R&D investment and scientific development. My personal blessing in this job is that I get to speak to our founder Bill Gates on a regular basis. He has this fundamental belief that most of the socio-economic problems can be solved by technological breakthroughs.
The particular cultural aspect (of Microsoft) I personally like most is its ‘tenacity’, with the belief that if you can focus on building technologies good things will happen.
Where do you see the improvement in the next 10 yrs? Is it in the mathematics? Or in algorithms? Or is it the computing power?
We are looking for a fundamental change in the structure of the web. Because HTML is a hypertext, the physical structure of the web is in ‘links’. The fundamental information structure is in those links.
ADVERTISEMENT
They are the roads and the bridges that enable us to navigate the web. The roads and the bridges of today’s web are nouns and noun phrases. And that’s why our search engine model works because key words are also noun and noun phrases. There are no verbs. The web even today has far outgrown its intellectual heritage of a corpus of hyper-texted document. It is becoming a digital society, whereby every human being, every location, every physical product has a digital representation on the web through devices, through software.
Good news is, the structure of the web is going through profound, dramatic, rapid changes that are unfolding in front of our eyes.
Is it a natural change or a directed one?
The best way to predict the future is to build one. Companies such as Microsoft have a unique opportunity to push our industries collectively. I give you the example of Bing for TVs, where you can use voice and gesture to discover video content on the digital TV.
Today, all you can do is use the remote. What if we could enable each to say ‘Oh, I remember my friend told me about a new movie.‘ Or, ‘there’s a TV series about that particular story”. The Xbox Kinect device has a camera. And our software can detect finger-level movement.
ADVERTISEMENT
As long as you know how to speak and wave your hand, we can empower future digital living room experiences through software and services. We are building computational understanding of consumer needs and interests to bring rich content into TVs. So till search reaches a certain level of sophistication, can you use manual curation like Wolfram Alpha?
We use curations already. Algorithms are overglorified. All those algorithms are a chain of human judgment data. So to your question, can we use more curation, absolutely.
A lot of differentiation is in the clever use of curation. Wolfram Alpha is interesting. Stephen Wolfram is a terrific scientist. He is collecting factual knowledge as a basis and then using symbolic representation to answer questions and to create new knowledge. And we have been collaborating with Wolfram.
ADVERTISEMENT
For certain queries we use Wolfram in the back-end. Sensible use of human curation can be a good way to differentiate our search experience. How big can it be? There’s obviously a limit to how much you can curate.
ADVERTISEMENT
People are not paying attention to crowd sourcing. For example “the EXC games” designed by a professor at Carnegie Melon. Essentially, it’s a game for labeling images. You randomly pick an image and you have to describe what the image is. If you and I agree we score points. The end result is you get more and more points. So humans start to curate images. And these images get into products and algorithms. But we can have different ways of designing those games, different ways of crowdsourcing so that human activity leads to more production of digital products.
Download
The Economic Times Business News App for the Latest News in Business, Sensex, Stock Market Updates & More.
The Economic Times Business News App for the Latest News in Business, Sensex, Stock Market Updates & More.
Related Articles
Building a frontier ecosystem over models essential for global economic value: Microsoft CEOADVERTISEMENT
