Channel 9 - Discussions by kevinsch

Coffeehouse - Does anything from MS Research ever get shipped?

kevinsch — Sun, 11 Mar 2007 05:12:13 GMT

Stuff that got its start in MSR ships in products all the time.

Most of it isn't "new products" per se. MSR really isn't in the business of creating new products. MSR's business is to broadly advance technology by solving hard problems and then to work with MS's product groups to get those technologies into customers' hands. So it's a broad technology investment, not an investment in specific products.

That said, occasionally MSR does incubate a new product. Tablet PC came out of MSR. As did SPOT. A lot of the company's early interactive TV work was also done in MSR. There are a handful of other examples, and we have a couple of other incubation projects in the works that may one day be introduced as new products.

Coffeehouse - C9 Gal Yay or Nay!

kevinsch — Fri, 13 Jan 2006 00:05:38 GMT

Yay, and she should be designed by a woman.

Coffeehouse - Microsoft BitTorrent

kevinsch — Wed, 22 Jun 2005 23:20:02 GMT

I beg to differ --Channel 9 rocks, but http://research.microsoft.com is a great and direct way for you to find out all about what MSR is doing. There's a whole "downloads" section with all sorts of cool things. And (fanfare please) RSS feeds for news, downloads and publications so that you can find out when something new just came out!

Coffeehouse - MS Research might speed up P2P

kevinsch — Fri, 17 Jun 2005 16:55:59 GMT

well, to answer the original question... it's still a research project so we can't say for sure if or when it will become part of a product, but we are actively talking with a variety of product groups about possible uses, and there's a fair amount of interest.

Coffeehouse - Studs from Microsoft

kevinsch — Tue, 31 May 2005 21:00:36 GMT

This was from a local Seattle show called "Almost Live" which ran for about 15 years here. So needless to say, everyone in the sketch is a celebrity here in Seattle.

Almost Live was Bill Nye's big break, and also where he created the whole "Science Guy" thing. About every other week, he'd do a sketch with some fun application of a science principle -- all decked out in his trademark blue lab coat and "Safety Goggles of Science". He'd blow things up, build huge catapults, all those things that our favoriate high school science teachers loved to do, and the audience ate it up.

I miss "Bill Nye the Science Guy." For that matter, I miss Almost Live -- it finally went off the air a few years ago.

Coffeehouse - Microphone buzz

kevinsch — Tue, 31 May 2005 04:04:59 GMT

Hearkening back to my days of being an engineer in a radio station...

A large fraction of audio buzz in sound systems is caused by what's called "ground loop." The theory of grounding is that by connecting compnents to the ground, you establish a common "zero" across all components -- if you test across any two points on any of your ground cables, you should get zero voltage.

It's a nice theory, but in practice since there is resistance in all wires (and in the ground), there's always some amount of voltage difference through the grounding system, and if alternating current leaks into it, you get a buzz. The fact that the buzz you're hearing is in the 50-60 Hz range is a very good indicator that you have a ground loop problem.

If it is indeed a ground loop problem, that would require:

1. two electrical devices, both grounded (e.g. plugged into the house mains by means of a 3-prong plug). You could get this by having a PC and separately powered speakers.

2. The electrical devices are interconnected (e.g. your poswered speakers are plugged into the audio card in your PC).

3. something is leaking AC current into the ground wire. This could be the power supply from either piece of equipment, another faulty component inside the PC, or (unlikely) even another electrical device nearly plugged into the mains (like your fridge, or your stereo).

Ground loops are notoriously difficult to isolate. So here are some things to try:

1. Disconnect your speakers from the PC (and from the mains, if powered) and try listening to your PC audio through a pair of headphones. If the buzz is gone, you just found your loop. That doesn't mean that the speakers are necessarily the problem, it just means that they are part of the loop.

2. Disconnect every non-essential peripheral from your PC, including your printer and particularly everything that has separate power.

3. leave your computer going with audio playing (so you can hear the buzz) then walk around and unplug appliances one at a time until the buzz goes away.

4. If desperation is setting in, go to a hardware store and buy an adapter that turns your 3-prong plug into a 2-prong (groundless) plug. Try using it BRIEFLY on your computer to see if it makes the buzz go away. Be aware that it is potentially dangerous to run your computer without proper grounding, so you do this at your own risk and this is NOT a long-term solution -- just a diagnostic tool.

If none of this works, post something here again and we cain brainstorm some other ideas.

Coffeehouse - Robots to watch children showcased

kevinsch — Sat, 05 Mar 2005 23:00:24 GMT

Unfortunately, the article got the project's emphasis a little wrong.

The teddy bear is a toy. The research project is really about using vision, audio processing and robotics to take interactivity to the next level -- as you move around, the bear's gaze can follow you. As different people talk, it looks at the speaker. If someone it doesn't recognize gets too close, it mimics getting "bashful."

It's really not meant for monitoring/watching over your kid -- it's just a toy. It's not spooky, it's just kida cute.

The people who created it have also created a mobile robot that uses much of the same technology. That, they imagine, might roam around your house (or your office) while you're away and be a form of telepresence -- smart enough to look in the right place to interact with people it might run into, so if you're remote you only have to do a minimum amount of "control" of the device itself and can focus on interacting with people through it.

Coffeehouse - Please interview Doctor Eric Brill

kevinsch — Tue, 31 Aug 2004 04:27:31 GMT

We did, albeit briefly. Scoble and I ran into him in the hallway when we were doing the MSR walkaround tour -- he's in the section featuring Gary Starkweather and the hardware research lab.

Coffeehouse - Brain drain

kevinsch — Tue, 10 Aug 2004 06:31:15 GMT

A couple of things are happening here.

First, there is absolutely a downturn in the number of students in the US pursuing technical degrees. Americans are not encouraged to go into it, and post-9/11 it is much tougher for foreigners to get student visas to come study here.

Second, while the downturn is most prevalent in computer science, it turns out that there is as much computer science going on in American universities as there ever was -- it just ins't going on in the CS departments. It's happening in biology, physics, cehmistry, astronomy -- across all of the sciences, as well as the social sciences. Computing has migrated to the places where it is needed the most and will have the most impact.

Third, the upper middle class and upper classes in the US think of computing as a "trade school" thing, and actively discourage their kids from pursuing it. Tour the high schools in the Northeast -- the best computer labs and programs are in the large public high schools, and the prep schools have meager computer equipment. There is a strong message being delivered here. It's also nothing new: as a CS major at Dartmouth, I can tell you firsthand that it was class warfare.

Fourth, China and India have no such misconceptions, and are actively growing their computer science degree programs. And the growth rate is tremendous.

The bottom line: if computing moves offshore, and computing research follows, we have no one to blame but ourselves. It will be our own short-sightedness that got us into this. And since computing is now an essential part of all of the sciences, expect the rest of the sciences to follow suit. Outsourcing isn't a statement about cheap labor; it's a statement about the value that we place upon particular disciplines and whether we see them as integral to our economy, our society, and our future.

Coffeehouse - Microsoft Research Faculity Summit

kevinsch — Wed, 04 Aug 2004 06:35:14 GMT

It was a great event, if I say so myself. I'm very proud of my team for hosting it so well. Our relationship with research and teaching faculty is super important and we take it very seriously.

Coffeehouse - Would you guys interview the Machine Learning and Applied Statistics group at Microsoft Research?

kevinsch — Wed, 28 Jul 2004 19:32:43 GMT

It would be your own private safelist, not a public one.

Coffeehouse - Would you guys interview the Machine Learning and Applied Statistics group at Microsoft Research?

kevinsch — Wed, 28 Jul 2004 05:36:39 GMT

So here's a key insight as to how Microsoft's folks think about anti-spam:

There is no single technology or mechanism that will alone solve the spam problem.

We will end up with a chained series of them, with anti-IP-spoofing first, safelists probably second, and probabilistic spam-filtering last. Expect more stuff, like the computational challege, in the middle. Individually, they are all weak. Together they are almost unbeatable.

So yes, this is a case in point: computational challenges make life hard for legitimate mailing lists. But if your mailing list is in your safelist, then it bypasses the computational challenge.

Generally you should think of the ordering as deterministic first and proababilistic last (or least error-prone first and most error-prone last). But there are of course other complications, like the notion that you would want the first things to be low-overhead (overhead == CPU cycles, I/O, etc.) so that a large influx of spam doesn't become a denial of service attack.

And yes, all of your suggestions are good things to do too. But they also are not enough.

Also keep in mind that this is a war of escalation, and will not be won overnight. We're starting to make real progress, but we have a fair distance to go.

Kevin

Coffeehouse - Free as in speech

kevinsch — Thu, 08 Jul 2004 02:18:22 GMT

And ironically, Cory's speech was delivered at Microsoft, hosted by my group at Microsoft Research.

Actually it's not ironic at all. Because we're not afraid of discussion on this topic, contary to popular belief.

But I simply disagree with Cory here. Things aren't as black-and-white as he makes it out to be, and Microsoft doesn't get to choose which laws to obey and which to ignore. It has to obey them all.

Coffeehouse - Free as in speech

kevinsch — Wed, 07 Jul 2004 04:43:39 GMT

Wow, Jamie, there's a lot of stuff in there that borders on consipracy theory.

First of all, MS doesn't believe that one license fits all. We use a bunch of different ones. If you look at the stuff that MSR makes available, much of it is under a BSD-style license, including code -- which is arguably much more "free" (as in "no ball and chain") than GPL or other viral licenses that make you post code. The Visual Studio team does something similar with some components. There are some situations in which shared source is appropriate. MS has other licensing schemes we use for governments who want to have very broad access to source. The point is that we take a look at the situation for which we are trying to craft a license, and we find (or create) the most appropriate one we can. And yes, a lot of those keep our intellectual property right propietary.

Here's another way to look at it: Microsoft has spent almost 30 years and tens of billions of dollars to build a business that respects the intellectual property laws as they are written. And that's not just creating our own IP, but paying real money to license others' IP when it was required for the products we wanted to ship -- because that's what the law says. How would you feel if someone ripped that out from underneath you? Oh, I'm sorry, but those intellectual property laws... they don't matter anymore. Yeah, all those years that you followed them, sorry about that, your loss. What does it say about a country's commitment to a supportive business environment if it arbitrarily and capriciously abandons the economic infrastructure that domestic companies employing hardworking Americans used to create and preserve a thriving business? You can't just throw away intellectual property law.

Now let's separate patents and copyrights. There are certainly issues with the patent system today, and undoubtedly a lot of bad patents are getting issued. Microsoft has come out strongly in favor of patent reform. We don't think patents themselves are bad; we do think that poorly constructed, and poorly examined, patents are bad for everyone and hurt the system. But that is a problem that extends far beyond software patents; there are good and bad patents in every field, and they all need appropriate rigor. I have yet to see a convincing argument that there is a distinction to be made between software patents and other kinds of patents,and I believe that it's a red herring that distracts from the real issues of good vs. bad patents in general, and good vs. bad examining.

On copyrights: first, I need to state that I'm not a lawyer, and I do not have a deep understanding of copyright law. I also don't speak for Microsoft on this topic, so what I'm going to say here is my own naive understanding and my own opinion. That said, the issues with copyright and DRM basically boil down to "DRM applies very specific verbs to things that never had that kind of granularity before, and gives the copyright owner the ability to control different permissions on different verbs -- and that is eroding citizens' fair use rights."

First thing first: I went searching one day for a clear definition of fair use in the written (federal) law. I found lots of uses of the term "fair use" but I never found a definition. The real definition is being created through case law in the courts, in a piecemeal and often conflicting manner. I submit that it's a real problem that fair use is being defined by precedent and case law, and not by our legislators.

It is absolutely true today that a PC could be used as a tool for rampant piracy in violation of copyright laws. If Microsoft doesn't take reasonable steps to stop piracy, it will absolutely get sued by every major media firm out there. Microsoft has to obey the laws as they are written and interpreted today. If we don't like the laws, we can get Congress to change them. But you can't ignore them. Microsoft is really between a rock and a hard place on this: it wants as always to create great software that empowers its users; but it has to follow a set of ambiguous copyright laws written for a different age.

One last point: why do you believe that it will make Microsoft's developers happy to have more than one Windows? One of the biggest values of the patform is that there is a cnsistent API across Windows installations that makes it easier to write an application to a big addressable market.

Coffeehouse - 3D desktop

kevinsch — Sat, 03 Jul 2004 19:18:53 GMT

Actually, the most difficult part of 3-D environments is the input device -- or lack of a good one.

Just look at videos games -- you quickly get a good idea as to how complex they can be. Personally, I could never get the hang of the input model for Halo, and finally decided that I didn't want to invest that much time just so I could shoot things. Most other games degrade it down to essentially "2-D with jumping" to get around that problem.

One of the cool things that the MSR team did with Task Gallery was to build a system that gave you all the spatial-memory advantages of a 3-d environment while letting you navigate it with a 2-D model.

But it's a super hard problem, and a place where the needs and constraints of information workers and consumers diverge. You can't expect to have a game controller plugged into an information worker's desktop machine.

Coffeehouse - Will the future of digital photo display be in Longhorn?

kevinsch — Sat, 03 Jul 2004 06:16:32 GMT

It depends who initiates the conversation. If the researchers do, they'll be told to go away.

On rare occasions, a product gorup will come to us and say "our alpha testing is showing that X feature is missing, or if it just did Y then it would really be a killer product. Can you help us figure out whether that's possible?" And we do our best.

Remember that the researchers are all shareholders too, and they equally want to see the product ship on time with great features and great quality. Once we help them to understand the development cycle, they gneerally respect it.

Coffeehouse - Will the future of digital photo display be in Longhorn?

kevinsch — Fri, 02 Jul 2004 04:31:09 GMT

Funny, Scoble and I were talking about this very issue when he was interviewing me yesterday.

We're definitely talking to the Longhorn team, and they have access to all of the designs and code behind this prototype.

But part of building the kind of relationship with a product group that enables tech transfer is to understand how to respect the roles of the research team and the product team. In the end, the decision about what ships in the product is the product team's alone. They have a very difficult job balancing out features vs. risk vs. schedule vs. stability, and they are the experts at that -- researchers aren't. In MSR we completely defer to them, and we also defer on the timetable for announcing which features and technologies are in or out of the product. There are a set of decisions that they need the flexibility to revisit late in the game if necessary, and announcing too much too early takes away that flexibility.

That's a very long-winded way of saying "I can't say" because it's neither my decision nor my place to speak to it.

Coffeehouse - Is Microsoft going to make Smart Phones with bone-conductive microphones?

kevinsch — Fri, 02 Jul 2004 04:13:19 GMT

For all the people who haven't read the article, it's actually a combination acoustic and bone-conductance microphone, which lets us get a high fidelity signal (like an acoustic mike) while only picking up the speaker(like a bone-conductance mike.

We're still fidgeting with the noise-cancellation a bit (read: it's still a research project). We're talking to product teams, but it's still too early to say exactly how this will get shipped.

Kevin

Coffeehouse - Research at Microsoft

kevinsch — Sat, 15 May 2004 06:49:31 GMT

Funny you should mention that... my team in MSR is in fact making a video about how we do tech transfer. I'd be happy to do a video for channel 9 too.

And as for the comment about things going live:

WWMX
Netscan
TerraServer

all came out of MSR.

Coffeehouse - Research at Microsoft

kevinsch — Sat, 15 May 2004 06:41:51 GMT

OK, I'm here now. It's amazing what you find when you look at referrer lists on your blog...

We've had a ton of MSR technologies get transferred into MS products. Including the NLP technology; in fact, the grammar checker, SMartTags and language detection on Office are all based on the MSR NL technology.

Speech technology has gone into Windows and Office. Graphics technology has gone into DirectX, XBox, and some specific games. Networking technology has gone into Windows and XBox Live. Windows Media codecs for audio and video use technology from MSR. Lots of other stuff too.

Some product groups actually started in MSR, including the Digital Media Division, Tablet PC, and SPOT.

Here's a nice list. There's more too.

Some of the development tools that were created in MSR's Programmer Productivity Research Center are now in widespread use throughout MS's product development teams.

It's not an exaggeration to say that essentially every MS product today has MSR technology in it.

We are always looking for new prototypes and code that we can release on the web site. Right now we're working on RSS feeds so that it will be easy for people to find out when we post new stuff. Stay tuned.