11:12 AM
FutureMan
P Anandan, a computer vision imagery scientist and founder MD of Microsoft Research India, feels that India’s image as an IT services hub will soon metamorphose into that of a R&D powerhouse. He spoke to DNA about the mission and challenges ahead of Microsoft Research in India. Excerpts:
Is it just for talent that you have set a R&D centre here for or is it that India itself throws unique problems because research does not exist in vacuum?
The answer is both. For research to succeed you have to have talent. It’s also a motivational factor working with best scientists and talent. Secondly, being in India draws our attention to some of the problems which you would miss elsewhere.
Just to differentiate, my lab only does research and no development. We have 60 people. The India Development Centre in Hyderabad has 1,500. Most of our staff are PhDs. By the time you get a PhD, you probably forget how to write good programmes. We work with our partners in Hyderabad and the US to do development, but we don’t do development ourselves.
Example of some of the Indian problem based research you undertook…
Interestingly, you will find that they may not be specific to India but were motivated from India. For example, most of the times the road traffic in India is a nightmare. Our researchers came up with an idea to monitor traffic in Bangalore. We thought how we can create an infrastructure that allows information from different cellphones and then analyse and based on that we can detect where the traffic is. And maybe we can provide that information to people who will be the pre-informed about the condition of their respective routes. For instance, if a cellphone is not moving for sometime it may be stuck in a traffic block and so on. Or if you hear lot of honking and noise then you may assume traffic blockage etc. So, information is there.
How does research evaluation happens in Microsoft?
When I evaluate my researchers at the end of a year I ask, well what have you done, well I have presented four papers in this conference and one of them has got best paper prize and my work is used in a course in MIT. That’s good. Then if someone says I tried publishing six papers and got zero acceptance or has got one acceptance and even the one that has been accepted is not of good quality then I realise this researcher is not doing well. So, we really expect them to prove themselves in the research community. In order to do it they will work on problems that research community feels is relevant and lead the community on the problem. That’s where the motivation comes.
Innovation is meaningless if it does not reach its user…
That process may take a long time. You don’t evaluate it in a short period of time of say anything from one month to ten years. For example, when I was working in the US lab I had worked on a technology to take images and then creating 3D models on it..3D just from two images. This was going on from 1995 to 2000. Our group was the pioneer in the world in that work. At that time Microsoft was not too interested in 3D imagery. So it was just sitting there. But just four years ago the Bing Maps team decided on 3D reconstructions of buildings and scenes of the world. So, they created a product called Photosynth. You can take your picture and it will give you a 3D imagery of the picture. So, it sat for six years without any commercialisation. But when the management thought about something like that it was readily available.
How does your research output get assimilated in products?
We have a special group called advanced development and prototyping. Their charter is to constantly listen to the product groups and our researchers for opportunities for interaction.
Which are some of the research areas you work on?
We have seven research areas. First is algorithm research. One of the reason we chose that as our research area is because worldwide algorithm is where Indian researchers are extremely good. We have a group which is working on multilingual systems.
What is the next step in multilingual research?
Ultimate goal is as a user you should be able to communicate with the machine in your own language.Trust me we are not anywhere near that.Another area we are working on is rigorous software engineering. We write many software programmes. We encounter with so many bugs. Similarly there are millions of programmes that might have encountered similar bugs.
However, every time a bug appears it comes up as something new. This is not documented properly. So, we are developing data mining techniques to address that. We would develop tools that help programmers debug. We would analyse the code automatically and would be able to say that here is the area in the programme which is not doing the right thing. We have a research group on cloud computing. We have a group called mobility networking systems that is looking at efficiency in mobile computing. We have a group working on technology for emerging markets.
Will you add more areas of research?
Yes, its fluid. For example, vision graphics is the latest addition. Tomorrow a researcher may say I am not interested in mobility I want to research into a new area. If the researcher is good and convinces us that he can attract other good researchers we go ahead. And if in an area we see only few are interested to research we would close that research.
How is your lab evaluated internally?
Our lab is evaluated on its productivity. On criteria such as are we recognised worldwide? Do our peers, say in MIT, Stanford and others, think that this lab is doing important work? When we came five years ago the number of computer science PhDs India was producing a year was just 35. And the quality was not that good.
IITs and other places were struggling to hire faculty. Because most people did not believe India was a good place to do computer science research. We were working with academic partners right from the beginning trying to evangelise the idea of more computer science PhDs. We gave PhD fellowships, travel grants to students, because only when you travel to international research conferences you are exposed to ways of doing research.
The other thing we are doing is using research community to work on projects of social importance. We are doing a project called India Digital Heritage project. We did a demo on how 3D technology can be used to capture heritage sites and tell narratives. The government’s department of science and technology (DST) got interested in this. Today we are looking at a national project to be funded by the DST involving about a dozen researchers from across the country.
Published: Thursday, Mar 11, 2010, 2:01 IST
By Amit Tripathi | Place: Mumbai | Agency: DNA
Is it just for talent that you have set a R&D centre here for or is it that India itself throws unique problems because research does not exist in vacuum?
The answer is both. For research to succeed you have to have talent. It’s also a motivational factor working with best scientists and talent. Secondly, being in India draws our attention to some of the problems which you would miss elsewhere.
Just to differentiate, my lab only does research and no development. We have 60 people. The India Development Centre in Hyderabad has 1,500. Most of our staff are PhDs. By the time you get a PhD, you probably forget how to write good programmes. We work with our partners in Hyderabad and the US to do development, but we don’t do development ourselves.
Example of some of the Indian problem based research you undertook…
Interestingly, you will find that they may not be specific to India but were motivated from India. For example, most of the times the road traffic in India is a nightmare. Our researchers came up with an idea to monitor traffic in Bangalore. We thought how we can create an infrastructure that allows information from different cellphones and then analyse and based on that we can detect where the traffic is. And maybe we can provide that information to people who will be the pre-informed about the condition of their respective routes. For instance, if a cellphone is not moving for sometime it may be stuck in a traffic block and so on. Or if you hear lot of honking and noise then you may assume traffic blockage etc. So, information is there.
How does research evaluation happens in Microsoft?
When I evaluate my researchers at the end of a year I ask, well what have you done, well I have presented four papers in this conference and one of them has got best paper prize and my work is used in a course in MIT. That’s good. Then if someone says I tried publishing six papers and got zero acceptance or has got one acceptance and even the one that has been accepted is not of good quality then I realise this researcher is not doing well. So, we really expect them to prove themselves in the research community. In order to do it they will work on problems that research community feels is relevant and lead the community on the problem. That’s where the motivation comes.
Innovation is meaningless if it does not reach its user…
That process may take a long time. You don’t evaluate it in a short period of time of say anything from one month to ten years. For example, when I was working in the US lab I had worked on a technology to take images and then creating 3D models on it..3D just from two images. This was going on from 1995 to 2000. Our group was the pioneer in the world in that work. At that time Microsoft was not too interested in 3D imagery. So it was just sitting there. But just four years ago the Bing Maps team decided on 3D reconstructions of buildings and scenes of the world. So, they created a product called Photosynth. You can take your picture and it will give you a 3D imagery of the picture. So, it sat for six years without any commercialisation. But when the management thought about something like that it was readily available.
How does your research output get assimilated in products?
We have a special group called advanced development and prototyping. Their charter is to constantly listen to the product groups and our researchers for opportunities for interaction.
Which are some of the research areas you work on?
We have seven research areas. First is algorithm research. One of the reason we chose that as our research area is because worldwide algorithm is where Indian researchers are extremely good. We have a group which is working on multilingual systems.
What is the next step in multilingual research?
Ultimate goal is as a user you should be able to communicate with the machine in your own language.Trust me we are not anywhere near that.Another area we are working on is rigorous software engineering. We write many software programmes. We encounter with so many bugs. Similarly there are millions of programmes that might have encountered similar bugs.
However, every time a bug appears it comes up as something new. This is not documented properly. So, we are developing data mining techniques to address that. We would develop tools that help programmers debug. We would analyse the code automatically and would be able to say that here is the area in the programme which is not doing the right thing. We have a research group on cloud computing. We have a group called mobility networking systems that is looking at efficiency in mobile computing. We have a group working on technology for emerging markets.
Will you add more areas of research?
Yes, its fluid. For example, vision graphics is the latest addition. Tomorrow a researcher may say I am not interested in mobility I want to research into a new area. If the researcher is good and convinces us that he can attract other good researchers we go ahead. And if in an area we see only few are interested to research we would close that research.
How is your lab evaluated internally?
Our lab is evaluated on its productivity. On criteria such as are we recognised worldwide? Do our peers, say in MIT, Stanford and others, think that this lab is doing important work? When we came five years ago the number of computer science PhDs India was producing a year was just 35. And the quality was not that good.
IITs and other places were struggling to hire faculty. Because most people did not believe India was a good place to do computer science research. We were working with academic partners right from the beginning trying to evangelise the idea of more computer science PhDs. We gave PhD fellowships, travel grants to students, because only when you travel to international research conferences you are exposed to ways of doing research.
The other thing we are doing is using research community to work on projects of social importance. We are doing a project called India Digital Heritage project. We did a demo on how 3D technology can be used to capture heritage sites and tell narratives. The government’s department of science and technology (DST) got interested in this. Today we are looking at a national project to be funded by the DST involving about a dozen researchers from across the country.
Published: Thursday, Mar 11, 2010, 2:01 IST
By Amit Tripathi | Place: Mumbai | Agency: DNA
0 comments:
Post a Comment