J_K9 @ Linux

Day 3: The Labs

March 24, 2007 on 5:09 pm | In Seattle 2007 |

Today (i.e. yesterday) was Friday, in case the times and date at the top of this post are confusing. We saw some amazing technologies and demos and we also learnt a lot in the process. We also had the chance to see Microsoft’s Saturn Lab, large server rooms with racks stretching out in long rows, and much more.

Presentations to Leadership Panel

Our first task at Microsoft was to present our projects to the Leadership Panel, consisting of James Pratt, a Product Manager of Windows Mobile, Jensen Harris, the General Product Manager of the User Interface section and Gray Knowlton, the Senior Technical Product Manager of Microsoft Office.

First up was my group with a presentation on a product called equate. We presented a strong case for a potential rival in the Point of Sale and Stock Management markets for small retail companies, such as local shops, which would undercut current solutions in many ways. The presentation lasted between 15-20 minutes. The Panel then had 10 minutes in which to raise any questions or doubts they had about our product. The same occurred for the other two groups, one of which was offering a PDA-based alternative to travel guides and the other an expanded version of the Mobile Suica technology. After an hour and a half of presenting and replying to questions (which was much more interesting than it sounds), the Leadership Panel went away to discuss our presentations and product to develop some feedback and also to decide which of us had presented the most viable product which Microsoft might consider investing in.

When the Leadership Panel returned, each of the three judges offered both kind and constructive comments on the different aspects of a project’s presentation, from the product itself to our presentation skills. Finally, after considerable useful feedback (which we all learnt from), they declared that the team with the best product of the three was the second group, who had presented a PDA-based travel guide system which was innovative and took advantage of some of the technologies which are beginning to be adopted by the general public. They said that the decision had been very narrow and that the other groups were not far behind. Oh well, as they say: better luck next time

Off to RedWest

After that, we headed off to Redmond West which is a smaller campus a short drive away from the main campus. This is where MSN and Microsoft Research are based, among other divisions. Our first stop there was the cafe where we grabbed our lunches. We moved to a conference room in a nearby building, where we were scheduled to have our next talk, and ate our meals outside our conference room. A few minutes after we had all finished and cleaned up, Drew, the Lead Scientist behind PhotoSynth appeared and led us into the conference room. I think he is the first one on the left in this picture of the team.

PhotoSynth

PhotoSynth is a truly stunning piece of software. Drew Steedly showed us some previously put-together collections as well as one which PhotoSynther created and compiled from a set of ordinary photos before our eyes. Drew also explained how PhotoSynther works, detecting edges and objects in images which stand out and attempting to match them to the objects detected in other pictures, as expected. It was all very interesting, but the final touch was to show us a 3D PhotoSynth Collection of one of our school’s courtyards compiled from images taken by about 20-30 different people which had been found on the internet. That left us all absolutely mesmerised!

I highly recommend that you try out the Tech Preview on Microsoft’s Live Labs website – it is worth the slight bandwidth usage spike. Live Labs is a sort of intermediate step between Microsoft Research’s internet-based ideas and production services, so you should expect to see PhotoSynth on public websites fairly soon. Wouldn’t it be nice to be able to take a series of pictures of your house and have PhotoSynther create a 3D image of it?

MSN SOC and Saturn Lab

We then moved over to MSN’s SOC, or Service Operations Center. This is the central location from which all events are tracked and dealt with. If a server or switch goes down, this is where it lets them know. If a service fails, this is where it lets them know. The events are also prioritised as they enter the system, allowing the SOC team to deal with the events accordingly. Most events have pre-configured responses which are called in a cause-and-effect way when a new event appears. If this response does not resolve the problem, the SOC team may attempt to fix the problem in another way and, if this also fails, the problem is moved up to Tier 2 where it can be dealt with by more experienced and specialised IT professionals. The lady who explained this to us was also highly enthusiastic and clearly enjoyed her job.

We were then given a tour around the Saturn Lab, a series of rooms containing server rack upon server rack. The Saturn Lab is used as a test system for products that likely later make it onto Microsoft’s MSN or Live production servers. Despite the relatively small size of the rooms compared to the server rooms of production environments, we were all amazed. I just felt sorry for the poor chap who had to do all the wiring on the end wall to connect all of the servers to the different network systems!

In addition to this, we visited Area 51. This is a highly important area which contains the core routers for a multitude of different servers and services. To give you an idea of its significance, it has a cooling system AND a backup cooling system which kicks in as soon as the first goes down. It has its own generators, which also power the SOC rooms. The building is built to be slightly mobile and thus is resistant to earthquakes. This should all be implying something about the areas we visited

Microsoft Research

We then moved over to another small campus to visit one of the Microsoft Research departments. We were originally scheduled to see Andy Wilson, Ph.D, but as we had arrived twenty minutes late he was unavailable for a short period of time. However, we were able to speak to a man called Chris Wert, who had done some interesting research on language translation tools, in the meantime. He explained the techniques used by translation tools and answered our questions on where translation tools will go in the future. From a personal point of view, despite being a linguist I would not like to work on translation tools – it seems to be an incredibly difficult task with an infinite number of possible improvements. The computer does not understand: it matches a certain sequence of words to a pre-learnt sequence and provides the most relevant translation for it. For example, there is an indefinite number of idioms in all the world’s languages, all of which could not possibly be added to the recognition system because some cannot, quite simply, be translated from one language to another because they lose their true meaning. Despite the drawbacks of working on a translation system, Chris seemed to have faith in his work and I wish him the best of luck in the future.

Andy Wilson came by not long after and showed us a video of some of his work over the past few years. Again, this was amazing: one idea led onto another and, while their practical use might not be immediately apparent, they were very cool to see in action. From a business point of view, these “useless” inventions might later lead to an idea which is can be manufactured on a public scale and marketed accordingly, so I do not think that his efforts are wasted. I’m not sure how much of this technology I am allowed to talk about, but I am absolutely certain that the final work he showed us and allowed us to play with, one of his latest works, is not covered by the NDA.

What was this? A hacked Xbox connected to a prototype webcam, a projector and a screen. This was truly unbelievable. The screen was sitting on a desk to the side. In the middle was a large, flat white desk which had nothing on it. Suspended a metre or so above the desk were a project and a webcam. However, this was not an ordinary webcam – as well as producing an ordinary image of the table, it also had a ring of infrared wave emitters and detectors around the edge of the camera which are able to detect the distance of a certain object by performing nanosecond calculations of the time taken for a certain wave to be emitted and received as it reflected off a particular object. Very cool.

When he fired the Xbox up and turned on a controller, the screen turned on and we saw a view from the webcam’s perspective of a small 3D blue buggy. The projector simultaneously projected a 2D image of the same car onto the white desk in exactly the same physical location as the buggy on the screen. He then placed an ordinary piece of card, folded in two, on the desk and it created a small mountain of the same height and edge gradient (steepness) as the card on the screen, which the buggy could drive up and fly off (which would be depicted as such both on the screen and on the white table by the projector). Turning on another controller caused a red car to appear in the middle of the world (both on the screen and the table), and adding another card caused another mountain of a different gradient and size to be created. You could also change the view mode of the screen to allow you to see the buggy and world from a different perspective, such as a tracking camera directly behind the buggy. Again, almost unbelievable and VERY cool. Finally, connecting your thumb and forefinger alone would create a box/crate in the world, as the connection and circular enclosure would be detected by the camera and interpreted and reacted upon accordingly. On the technology side, it uses XNA for the on-screen rendering and some C++ to allow the Xbox to interface with the webcam.

Wouldn’t it be great to be paid to have fun?

Relaxation Time

That’s it for another great day at Microsoft. They certainly have some excellent engineers – I can’t imagine that creating PhotoSynth or that buggy invention are simple tasks (the buggy invention took Andy 5-6 weeks, although he had been working on it as a side-project for much of the time). We were meant to go back to the hotel and then to watch the Seattle Sonics play against the Minnesota Timberwolves (i.e. a basketball game), but so many of us were still jet-lagged that, to my dismay, we decided against it and went out for supper at a Japanese restaurant which offered a delicious buffet.

I will try to upload some pictures (such as of that buggy creation) later today, as I still have not had the time to buy the required USB connector to transfer the images from my camera to my computer, so watch this space!