Meet Microsoft, the world's best kept R&D secret
pcworld.com
pcworld.com
Two words, Xerox PARC.
At Sun there was a weird joke that Sun Labs was where good ideas went to die. It was frustrating.
The point here is that good R&D is a necessary but not sufficient component of innovation, the second is a willingness to productize your work. Strangely the hardest thing about that is not making a product out of it, the hardest thing is making a product you can ship.
Good R&D isn't constrained, which is to say that you don't tell the folks doing the research you are only researching things we can sell for a profit, but that is a constraint on products. What happens is the 'Apple effect' where you have a bunch of researchers who can't make a profitable product (Xerox Star) and then a product guy comes along (Steve Jobs) who sees the essence of the innovation, and can strip away the parts where it goes too far and ships that.
Its really challenging to build something close to your vision and not ship it, it seems like it is impossible to build something that is close to your vision and then ship something only half as close as that. But that is where the success can be. "Fumbling the Future" [1] is a fascinating read for that reason.
[1] http://www.amazon.com/Fumbling-Future-Invented-Personal-Comp...
Nokia developed HW, Symbian, S40, MeeGoo, and services. It also spent a bit on actual research. Pure research often pays off (in unpredictable ways). Developing a number of platforms turned out to be a bad idea.
S40 is still their monumental achievement:
You have made sweeping and broad overgeneralizations that betray the complexity of the situation. Apple's masterstroke was not in inventing the concept of a smartphone, but in producing one that provided a much better user experience that the rest of the market. As for the differences in hardware and rf engineering, I'll leave that to someone more qualified to explain because my knowledge in the area is far from complete.
Actually Nokia date back to 1865 but they originally a wood pulp mill then rubber manufacturer. They got into electronics in the 1960s and focussed on phones from the 1990s.
And if Microsoft Research is like IBM, then it's also only partly funded by the corporation, and regarded as a semi-separate entity. I think they see it more as a contribution to science and as a long-term investment rather than product research.
The second time was about three years ago during St. Patrick's day at the Hilton. I was able to use most of the applications described above.
In a sense we are product focused as well, except our target market isn't consumers, but rather people up on high within microsoft. We have to sell our research to them one way or another, and it's often the research teams that have the best advertising sense about them that end up doing the best.
MS seems to have some other problems, productizing their research is part of their problem, in which case they are likely spending way too much on it. Another, I'm gonna go ahead and say it since it keeps coming up here in different capacities re: IE10 and Windows 8 and other things, they have negative brand equity with the real early adopters and leading edge geek folks due to their years of treachery and just not being nice. While kinnect is cool (very very cool, my 2 year old can play games!) I think I'll wait for Google and Apple to solve the wearable ultra personal computer problems before I'd let MS in to my personal space with some of those phone gadgets. Their final problem, IMO, they still sort of don't get it. Take JPEG-XR, it's a nice piece of work, they even played nice with ISO and it's a standard of sorts; good on them for it but why no BSD or Apache licensed free implementation with a written down patent promise? They planning on drawing funds off that patent? Or would they like digital cameras to start using that stuff? Mono is nearly 8 years old and it's sort of used but largely ignored due to MS and potential patent problems, I think they are very solvable problems. They just don't play well with a lot of others and that kind of undermines the research, it doesn't matter how cool the invention is if you have to eat the rest of their shit with it.
> it doesn't matter how cool the invention is if you have to eat the rest of their shit with it.
But it does matter, because even if MS doesn't pick it up -- or implement it well -- someone else might.
Unless there are potential patent problems.
not just a willingness to productize, but the ability to do it successfully. Kinect, for example, is obviously a huge technological accomplishment that is the sum of a ton of research and the work of many smart people. but it's just a gimmick for a game console, that many gamers are pretty ambivalent about. I can't help but think it's a bit of a waste of talent.
I'll be interested to see if there has been much improvement in the technology when the next Xbox launches this fall.
This is why I believe Apple are holding onto so much cash. Far easier to ask banks for million dollar loans when you have billions in savings rather than begging private equity who'll strip you to pieces.
It seems like their problem is that actually that they are not bold enough to push some of their research ideas and prototypes to the market. Why do they need to copy Apple or Google in order to succeed in mobile market? Why not to do something bold and push it?
meet dynamics crm, sharepoint, biztalk et al. (and f*in shoot yourself)
This is the core dynamics behind lot of large companies with gigantic R&D without focused goals. And it also very well may be the reason why they contribute marginal to company's bottom line and still stay alive year after year. Like Nokia's R&D they are usually first one to go as the company's revenue starts declining.
Bear with me, they've effectively laid out all the pieces:
- A "omni touch" interface that allows for interaction without physical input devices.
- A variety of leaps forward for Kinect to map and translate your physical environment to a data stream.
- A "holodeck" and other tools that overlay interface design onto physical objects.
- Foveated 3d graphics and other leaps forward that would drastically reduce hardware costs for graphic rendering.
Nearly every one of these projects are directly applicable to the "wearable computer" concept.
I wonder if, looking back in 10 years, we'll see that MSFT's currnet weak smart-phone entries were ultimately not that important to their overall strategy.
My issue with this article is that it forgets about the ugly side of Microsoft: delaying the sale of innovative new things to protect old profit centers in the enterprise market. Does anyone really believe that Microsoft CAN'T make editing hosted documents in Sharepoint as seamless and smooth as Google Docs? I truly believe that they have chosen to neuter a lot of their collaborative, web app based software to keep their old trusy desktop software (Office) profit center alive and kicking. They benefit from this, in the short term, in two ways: If the latest version of Office has new features, they sell the upgrades and make bank. Because it's desktop, they can say "you need Windows 8 (or whatever is latest and greatest) to run it Mrs. CIO." So then they make the OS sale too.
Historically, Microsoft R&D has come up with some great stuff, it just really hasn't made it into their user products. Bill Gates was hammering on tablets and touch interfaces for years, but Apple made it happen. They have had great microkernel OS or modern language OS R&D projects, yet the Windows kernel hasn't changed dramatically. There is a missing step between R&D and revolutionary product improvements.
The title is misleading. The work described is done at MS Research, the relation of which to corporate MS is not quite clear.
Many companies like to fund R&D divisions as a means to attract the best minds from the academic community in the hopes that they will benefit from their talent in some way. It's often not directed work but more of a recruitment tool. The best professors bring with them the best students and many of those students end up working on products not just research.
However, in my time at Microsoft I worked with the research team on a couple of occasions trying leverage their ideas into real products with little success. We would send them some interesting problems (in one case we asked them to spend some time on snow accumulation algorithms for a snowboarding game) and they would disappear for months and return with a cool demo that was impressive but usually failed to meet the given criteria that allowed it to be used in a shipping product. e.g. speed, memory efficiency, data size requirements etc.
What precise things were problematic? Their disappearing for months? Their not communicating enough? Them royally ignoring practical guidelines from you re: speed, memory etc?
More like Microsoft wastes more money on R&D that doesn't convert in revolutionary products more than both Apple and Google combined.
We've always known Microsoft spends a lot on R&D and they like to make those "20 years from now" videos, but I haven't seen much come out of it. Last year they even bragged about how they had the idea for the iPhone 20 years ago. But so what? How did that help them? At best it helped them create Windows Mobile and the PDA's 10 years later, but that was a niche market, and Microsoft never had much market share with Windows Mobile in the smartphone market, which was a lot smaller than the current smartphone market back then.
So I guess the moral of the story is that "lab inventions" don't mean much, and you could waste a lot of money on them, and with very little to show for it in the market. I'm pretty sure the $2 billion dollar Kin project was part of that R&D spending, too.
It is actually quite known in the tech industry (among those who really care about technology and not just business) that after the slow decline of Bell Labs and Sun Labs, the two most innovative tech companies are IBM and Microsoft. Actually, I've always assumed everybody knows this.
I don't know if it means Microsoft (or IBM) "has a future", but who cares? There's so much good stuff coming from those two companies (or their research divisions to be exact), so why not just enjoy it?
Also, I don't know how it works at Microsoft, but at IBM Research, the company is responsible for only half the funding. The researchers have to secure the rest themselves from grants and such, just like in any academic institution.
This sentence is weird to read here.
We have a big company spending money on science that is not tied to what they can sell. That is a good thing. I am disappointed that anyone on HN can say anything different.
Of course, we could have a discussion about the difficulties of running pure research alongside a commercial arm. But that's not what your comment is doing.
“It embodies what we believe in and our sense of values,” Rashid said. “It states, first and foremost, that our goal as an institution is to move the state of the art forward. It doesn't matter what part of the state of the art we're moving forward, and it doesn't say anything in that first part of the mission statement about Microsoft. It simply says our goal is to further research in the fields in which we work.
“The second part is equally simple: When we have really good ideas, our goal is to make sure we get them into use, that we change the world with our ideas and with our technology. We take our best ideas and push them into our products as rapidly as we possibly can. That's important.”
Microsoft sponsors many top people in computer science and allows them to spend time on pure research and to publish academic papers. Ever heard of Simon Peyton Jones or Sir Tony Hoare?
How can that be bad?
For example, Microsoft pays most of the top Haskell developers to research and develop Haskell. There's no patents on that, and the source code and papers are all freely available. Microsoft isn't turning it into a product, it's just doing it to better computer science, to increase their standing, and to inform their other products such as C#.
And you say that's bad? Fuck you.
It is true that a portion of the research sponsored by Microsoft is published as academic papers, but I'd suspect that, in relation to the vast amount of money Microsoft is spending on R&D, those papers represent less than 1% of the results. We never hear about the hundreds of millions spent where the results never saw the light of day.
These accusations are false. MSR does not generally patent their work. I'm in language research and have not encountered any of their work that they've patented. Also, MSR's researchers and evaluated and appraised based on their publication record, so the only incentive is to publish. Microsoft's people publish a huge amount. Most university researchers are envious of how much they manage to publish, which they achieve through very high quality people, properly compensated, with lots of resources and being left alone.
Why do you "suspect" that only 1% is published? Where do you get that number from? Have you worked there? Have you collaborated with them? If they're only publishing 1%, the research would be superhuman.
You have put a false accusation to a group of people, and that pisses me off. It's not even like it's a difference of opinion - you've clearly not worked with these people or studied their publications, and are just making assumptions.
>>If they're only publishing 1%, the research would be superhuman.
Well yes, they spend over $10 billion on research. Of course it is superhuman. Perhaps you are lacking this context, which is why you find it hard to believe my suspicion that they publish a tiny percentage of all research findings.
As a matter of defense against patents, though, MSFT has a strong interest in insuring that its basic research is thoroughly documented in public, establishing prior art.
Of course transferring tech into products is part of their goal, but it's the product teams that continue from there (R passes the ball to R&D).
Regarding publications, as other pointed out in this thread, MSR does publish a huge percentage of their work.
I seriously hope you are a teenager. Because if you're an adult and you talk to people like that ... well then good luck in life man ...
Solved the meaning of life but posted it with speeling errers? We do still keep a few roadside crosses and nails laying around from old Rome; you'll get your comeuppance.
The idea is that MSR explores (R), then Microsoft integrates what they feel is interesting into products (R&D).
Research doesn't have to be sexy to be useful.
While designers have spent a decade gushing over skueomorphism, rounded rectangles and gradients, Microsoft has made two steps forward in the WIMPy UI development - the ribbon and Metro.
Furthermore, while Google and Apple are trying to partition the world into walled gardens; via TouchDevelop, Microsoft is the only major company pursuing mobile devices for general computing.
see plan 9.
Are you mad?!
In fields such as machine learning and languages, papers by Microsoft Research _dominate_ top conference proceedings.
It seems like their problem is that actually that they are not bold enough to push some of their research things to the market. Why do they need to copy Apple or Google in order to succeed in mobile market? Why not to do something bold and push it?
Even if you don't know of MSR, this should be of no surprise if you have any recall of the news you read over the years. A lot of the work they do has appeared in the NYTimes and such. For example, I remember very clearly that about 6 years ago, Bill Gates basically said that robots are the future. It didn't take me long to find articles and website to remind myself of what he said (see link below). Although Bill Gates is no longer CEO, he is still chairman and you can certainly bet that MSR will be leading the way in this research.
http://tech.blorge.com/Structure:%20/2007/01/07/bill-gates-r...
At least it's not just a marketing vehicle like IBM Research and they build real products like Kinect. But that was also a bought company afaik and didn't come out of MS R&D.
http://www.cultofmac.com/67951/how-apple-almost-got-microsof...
Just because it was purchased doesn't meant they haven't helped push the technology forward, either in pure R&D or by just promoting adoption.
- did the product design and put it into a friendly package (take apart a Kinect some day, they did some neat stuff in there)
- developed the body tracking algorithm (http://research.microsoft.com/pubs/145347/BodyPartRecognitio...).
- developed the microphone array, speech recognition algorithms
- provided developers the resources they needed to take advantage of the device
- applied it well to gaming and quickly brought it to market
- put the technology in the hands of millions, triggering even more innovation
- OpenNI requires a calibration pose, while the technique Microsoft uses allows players to walk in and out of the frame. Kinect would not have succeeded if people had to calibrate it all the time.
- Microsoft does predictive analysis so it's much better in non-optimal environments (e.g. a living room) or in cases where it loses track of a body part because it's behind something.
- Also. To train it. Microsoft generated and processed millions of depth images from mocap data taken of people of all sorts of different shapes in sizes in all sorts of environments. It's insane. And of course they needed new algorithms to manage that...
- Lastly - no audio/speech or 3D facial expression tracking in NITE
I bet a large percentage of the population in the world doesn't even know about Microsoft.
huge budget != disruptive innovation
Think about how most (all?) of the top tech companies you know about started: garages, dorms and kitchen tables. HP, Apple, Microsoft, Google, eBay, Facebook, Twitter, etc.
Sure, some needed huge money to scale, but the genesis of the idea-turned-product did not require millions or billions to generate.
Sometimes you have to wonder if hunger is far more valuable than money to spur innovation. I happen to think so. And, while I do not diminish the work being done at companies like MS, sometimes I feel that they are just throwing money at PhD's who are having a great time playing with interesting tech but simply don't have the drive, hunger and urgency to make something more than a great research gig out of it.
Research != R&D
Not trying to antagonize you, just pointing an issue that seems pretty common in this article discussion.
Maybe 8 or 9 years ago, I talked to the Dean of the University of Colorado Comp Sci department. I could only remember Benjamin Zorn as a faculty member, so I dropped the name. By then, Zorn had gotten hired away to Microsoft R&D. The Dean of the Comp Sci department made a number of references to Microsoft R&D as a "research roach hotel" - researcher's go in, but no papers ever come out.
If Microsoft R&D is a secret, then it's Microsoft's own problem, I assume.
Ad publishing isn't a panacea, there are many ways we should be communicating, if anything we publish too many papers as researchers.
http://research.microsoft.com/apps/catalog/default.aspx?t=pu...
That appears to be about four orders of magnitude more research than that used to fact check your story.
Using these ratings from 7 months of games from over 3 million players, we look at how play intensity, breaks in play, skill change over time, and other titles affect skill.
MSR encourages publishing almost to a fault. They even have a portion of their legal department specifically trained to support MSR's researchers so that the "corporate" tendencies from the rest of Microsoft don't bleed into MSR.
Example of the output: http://jeffhuang.com/best_paper_awards.html#institutions
Also look at any proceeding from conferences in areas where MSR works. Their output is pretty high (and high quality).
That said, if you ever become a PhD candidate in CS, you should apply to MSR for a summer. It's an amazing place to expand your network as a researcher.
This is my biggest beef with MS. They consistently wait for other companies to innovate, then wait to see if the market will support it.
Only when they see an advantage to developing their own version of a product will they enter a market. You can look at their MP3 player the Zune, you can also look at how long it took them to get into the smartphone and tablet markets.
I'm wondering if there are any concrete examples of them releasing a product which has utilized their arm of R&D for a specific product and pointed to it as a reason why it's available to the public. Similar to how Google always touts their R&D teams for developing a myriad of their products.
Microsoft certainly attempts to "productize", they simply fail on execution:
When the iPod was still "only" an MP3 player, Microsoft was pushing video capable PMPs through their PMC software.
Microsoft was betting on the tablet long before Apple.
Windows CE based Wizard style PDAs predate the Apple Newton, and Pocket PC was largely successful and shows Microsoft have been trying to get a Mobile OS into your pocket for nearly 20 years. It was largely Palm and Microsoft that built out this space.
The PocketPC iPaq even had a GSM jacket that allowed it to be used as a phone, making it a smartphone that predates even the earliest Ericsson/Nokia smartphones. Not to mention Windows Mobile which predates iOS and Android by many years.
Microsoft has been working on WebTV, Media Center and Xbox to get into your living room long before AppleTV or Google TV.
Microsoft has certainly attempted to productize on the ideas that come out of Microsoft Research and is almost always on the cutting edge of technology, and in many cases years ahead. The problem is that they don't have the same unified vision many other companies do, so when they execute they do so poorly: With incomplete products in an incoherent package.
Microsoft's problem isn't that they aren't making enough products, it is that they are sometimes too big and disorganized to make them good enough that people can see the underlying value.
Newton came out in '93 and WinCE came out in '96. But that nitpick doesn't take away from your general point. MSFT seems to be good at coming out with things that have potential, but then let's them languish. Their mobile products were a prime example. On the last WinMo phone I had (2006-ish), the OS had dialogs that were (I assume) user-drawn and going on ten years old. Those dialogs looked like it, too. Apple comes along with a platform they were willing to pour the company into, and MSFT has been playing catch-up since.
You say they don't make their products "good enough", I'd argue that Microsoft doesn't stick with their products long enough to make them good enough. Their wireless home phone system (can't find a link) was an example. V1.0 was rough around the edges, but useable and forward-thinking for the late 90's. v2.0 was going to be great, except there never was a 2.0.
I also think what you say about them failing to support their products in the long term is very true. Until recently, Microsoft has been willing to let a lot of its products languish both aesthetically and in terms of functionality.
I think this ultimately still comes down to Microsoft's large disjointed nature and stubborn commitment to support legacy hardware/software, which made it less dynamic than a company like Apple that had a strong top-down unified decision making process that allowed it to push out cutting-edge products and aesthetics.
It's weird a company have the funniest R&D labs and its main product - Windows - still sucks with 80' Windows Registry structure or a file system which fragments a lot.
All their amazing R&D didn't make W8, their phones, or their tablets any more enjoyable or purchasable.
http://www.forbes.com/sites/rogerkay/2013/01/02/microsoft-is...
Since most of these revolutionary systems aren't available as actual products, I have to assume that Microsoft buys people and ideas mainly to keep them from interfering with their monopolies.
What ordinary people mean by "innovation" is "what cool new things can be made available that noticeably change Grandma's life".
Most academic work is USELESS for this because it ignores too many constraints, making the market hostile to it. Richard Gabriel did a piece on this, called "Money through innovation reconsidered".
Don't buy academic hype.
How do you expect eg low-level brain research to be useful for your grandma? It can't, but that does not mean at all it cannot be innovative or that the researchers only care about the next paper.
Point1: Nothing that Microsoft ever does will blaze a new trail of products, they are not an original sort of company. They will always be playing catchup.
EG Kinect to WiiMote
Point2: Nothing that CS conferences or journals ever publish will blaze a new trail. They too will always be playing catchup.
EG first came Facebook, and only afterwards the whole social networking research craze in academia.
I exaggerate Point2 a little. There are 1 or 2 Academics who actually make stuff, like Michael Stonebraker. But overall the record is pathetic.
Don't buy the hype.
Actually, it was a lot of this published work that resulted in: Internet, Networks, Filesystems, OS(Unix), Compilers(LLVM) etcetc.
Granted, a lot of research does not result in any useful product, but that isn't what research is all about. Its about exploring the boundaries of our knowledge. Trying wild ideas that you couldn't do within the time and budget limits of a company... because some of those wild ideas do work spectacularly in the end
Those things are not exactly difficult to do at a basic level (the 0to1 level) although at an advanced level (1toN) they might be.
So you will find that how those things got onto the market was by some engineer figuring out how to do it and hacking it together, not the work of some professor.
The two examples I am familiar with are filesystems and OS. Our filesystems today are the result of descent from FAT12, which came from discussion between Marc McDonald and Bill Gates.
http://en.wikipedia.org/wiki/File_Allocation_Table
Our operating systems are the result of descent from CPM, which came from Gray Kildall deciding to write a RAM tester.
http://en.wikipedia.org/wiki/CP/M
And that applies to 90% of the market. "Look ma no research!"
I can find you similar examples for compilers and networks as well, though in those fields the examples will not be so outright embarrassing to academia.
What research is about is mostly chasing trends in conferences and funding bodies.
Don't buy academic hype.
That is a question without an answer. Probably, they would have been invented, probably not. The ideas themselves were played around and worked on by academics though. You can't deny that they helped to usher in the technologies much quicker than otherwise.
>So you will find that how those things got onto the market was by some engineer figuring out how to do it and hacking it together, not the work of some professor
Professors don't (usually) sit down and write code, they hire others to do it for them. As I said before, the point of research isn't to produce a product, ready for use. It is to try out ideas that wouldn't be tried by the industry. Its good that Marc Andreessen and Eric Bina were hired to code what later turned out to be the Mosaic web browser.
>Our operating systems are the result of descent from CPM, which came from Gray Kildall deciding to write a RAM tester. http://en.wikipedia.org/wiki/CP/M
MSDOS, true. Unix, false. Unix descended from MULTICS[0] which started much before CP/M[1]
>I can find you similar examples for compilers and networks as well, though in those fields the examples will not be so outright embarrassing to academia.
I wince at your choice of words. "Embarrassing"? Why? If the industry produces an idea or technique that solves a research problem, academics have no trouble accepting it. See for instance: how quickly AWS, CUDA etc are being adopted by academics for their research.
So, in total, you have given one "fact" which turns out to be untrue. And you say you can give more examples in networks and compilers. Please do. I am genuinely interested to see what it is that has so tarnished your views on the work of academia.
>And that applies to 90% of the market. "Look ma no research!"
Clearly.
Here, take another famous example: relational databases originated in a paper by an academic, E.F. Codd[2].
I see that you are hell-bent on showing that academic research has made no contribution at all to the development of modern computer science. I haven't got the time or inclination to argue, so this will be my last comment on this matter
[0]http://en.wikipedia.org/wiki/Multics [1]http://en.wikipedia.org/wiki/CP/M [2]http://en.wikipedia.org/wiki/Relational_database
Unix is irrelevant; nobody used Unix until Android came along. Unix is in fact the other 10%.
Furthermore, relational databases are the work by Professor Stonebraker, that I already conceded in my original post.
I haven't got the time or inclination to argue
No, what you lack is the time and inclination to apply reading comprehension to my posts.
-ex-microsoftie
Cheers,
BW
To learn that changing the GUI of Windows 8 was not that good of an idea.
Microsoft should have used R&D to make products more affordable.