This is pretty wild that some of the same coding techniques that we use to identify spam mail can also be applied to identify HIV virus mutations and this is a great way to use “machine learning” for better research and knowledge.  Today there’s so much code written that you can in fact see what parts or portions of existing queries, etc. are built in and use those for another industry outside of the original design.  Mashable does some good coverage on this and also you have to love the video when you look at how they are working with augmented reality and the race car.

Of course there’s no better device and technology than Kinect that makes a lot of this happen.  In the video they are using the “Beamatron” technology as it is called to control a virtual race car that sees stuff in it’s path and having a virtual toy car as such might put some of the other folks out of business, maybe?  They also continue on to the demonstration of using Power Point in the same fashion.  We all know that Kinect is coming to Windows and developers are hard at work to make more of this happen. It is very interesting to see a million dollar surgical robot working with the $150 Kinect device, link below at what Johns Hopkins was working with. 

Kinect And daVinci Surgical Robot Do Simulated Surgery Suturing Together (Video)

If that isn’t enough you can see another use of Kinect with radiology taking the need to “touch” out of the picture.  Last year I attended the Israel Conference and had quite a bit of time to speak with Microsoft/XBOX Corporate VP Ilan Spillinger and of course he was also very high on medical solutions with Kinect. 

Microsoft Kinect Working with a PACS Server-Images on Steroids Via Gestures (Video)

When you watch the video with the car and passing the virtual “ball” around you get all kinds of ideas on how lighting, or other information could be shared like even in an operating room or the ER for that matter.  BD

24 year Microsoft vet and Microsoft Research General Manager Kevin Schofield told me Microsoft Research is also focused, at least in part, on “solving big problems” like the HIV virus.

Medical Research would seem an odd fit for Microsoft, but the research group is filled with all kinds of experts, including a couple of MDs. One of them is also a computer scientist and became fascinated with how doctors make crucial decisions in high-pressure situations when they have incomplete information (think emergency room visits). This led to work in machine learning (Microsoft Research does a lot of work in this area), which uses what’s known to figure out the unknown. Spam filters work this way. They can look at email and if the word “Viagra” is in it, decide with some degree of certainty that it’s spam. Now, that research is being applied to HIV vaccine research.

The HIV virus is known for its tendency to mutate, which makes it hard for people to develop an immunity. A spam-like filter can find the known in the unknown — in this case the core, recognizable virus.

I asked Schofield, who began his carrier at Microsoft in the OS group, if Microsoft founder and former CEO Bill Gates used the legendary Bell Labs as a model for Microsoft Research. “At a 30,000 foot level,” Schofield said, “Bell Labs was the model, but on the ground it was more Carnegie Mellon.” There are, apparently, two styles of research lab: The Xerox Parc model, where you isolate research from the business and the other model, where you basically have business fund and drive research. Both approaches can hinder tech transfer. The second method, in particular, said Schofield, tends to guide research too strongly so that most of the work is spent delivering product enhancements. Schofield said Microsoft took the middle road.

Like any true research facility, Microsoft Research has its share of duds. I asked Schofield if there was a room where Microsoft Research keeps its failed experiments. “Failures would be a big room,” Schofield said, laughing. But he also counts some of those failures among Microsoft Research’s successes.


Post a Comment

Google Analytics Alternative