Slashdot Mirror
Mitsubishi Breaks Up Famous Computer Science Lab
Andrew Koyfman writes "Mitsubishi Electric Research Laboratories is falling apart. Top researchers and scientists are being poached by the competitors, including BAE, Adobe, and others. The lab was responsible for much breakthrough research in the areas of computer vision, computer graphics, AI, and machine learning. They were the first group to develop the Diamond Touch table, an early precursor to Microsoft's Surface Computing. Now it looks like the famous lab will be no more, at least not in their original glory."
suggests the Microsoft table is a vast step forward. Hardly!
Engineering is the art of compromise.
that a number of char. system said that the new fangled gui system was a passing fad. And then a guy from silicon valley got a free pass to see it at Xerox. Later,these guys passed it on to a small company who was doing compilers for them. From redmond. There, the small company was given ALL sorts of insider info. And what was not given, was flat out stolen. And it all became a vast step forward.
Ms said that the table was a vast step forward. They did not say that they developed it. Just that they are going to build them.
I prefer the "u" in honour as it seems to be missing these days.
There's really no surprise here... if a lab doesn't spin off something valuable or at least has something big in the pipes that could be marketable in a a few years - cut your losses and shut them down.
Coming up with a table computer is really not cutting edge - even before Microsoft stole the idea.
Including this.
And this and this.
And this.
This article is a CLASSIC example of why we need academic institutions and government funded research into the great unknown.
A significant number of these types of labs, doing pioneering work under the name of a parent company, have been gutted with the intent of making them more "product focused" with the intent of converting brain power straight into $$$.
Well, here's a news flash - that's not how real research (as opposed to product development) works. With research into new stuff, you DON'T KNOW what you will find or what it will be worth. NO ONE does, BY DEFINITION.
A corporation can only do this type of work when a) they have a decades long focus and b) have sufficient profit margins to soak up the cost of research without immediate returns. That's a rare situation, and it's becoming rarer in a more competitive world economy.
Rather than bemoan this behavior (after all, money making is at the heart of commerce) we should be funding basic research at universities at much higher levels. Funding at universities has gotten tough enough that they will undertake a wide variety of investigations for commercial companies just to pay the bills. This makes them de-facto corporate research labs, and takes away time from their exploration into the unknown. Grad students become extremely cheap labor for companies, just indirectly.
Right now, it won't matter commercially. Product cycles don't get impacted by long term research for years or decades, so for a while we won't see this problem. But it's going to hurt us in the end. As products stagnate, foreign plants will catch up and learn how to produce at higher quality. They will begin to match or even exceed the performance of existing outputs domestically, and we will not be able to compete because there will be nothing in the long term pipeline that might convince people to stay with us.
Pure Research HAS A POINT. Even if the profound social and philosophical questions surrounding the pursuit of knowledge for its own stake don't register, it can also be viewed as a long term investment in our future. Balance sheets and profit statements do not define the whole of human existence, nor do the look far enough ahead to see long term consequences.
Again, it is unrealistic to expect this of businesses - that is not how the system is encouraged to behave. However, the government SHOULD be thinking about these issues. They need to be funding a LOT of basic research into all manner of alternative energy science, and the more basic science behind it - and thats actually a more practically centered goal. Truly BASIC research into the unknown, with no end game in mind, seems to be a tough sell nowadays.
Corporate research works ONLY when the long term is viewed as Very Important. It's dangerous to trust to that in an uncertain amd extremely competitive market.
In an era when Nintindo has passed Sony in market cap, it pays to focus your research efforts on areas relevant to your core competencies rather than blue-sky research into market segments where your presence is negligable.
Hell, even the classic example of Xerox PARC is one of a brilliant organization whose parent company was woefully unable to commoditize the ideas there (their GUI licensing deal in exchange for Apple stock is among their few commercial successes).
Publically held corporations exist to make stockholders money, not to do research "because it's cool." Period.
Lawrence Person (lawrencepersonh@gmailh.com (remove all "h"s to mail)
http://www.lawrenceperson.com/
What was the name of that old Dig-it-all labs site that Compaq and later, HP disappeared?
No folly is more costly than the folly of intolerant idealism. - Winston Churchill
In the USA, the only industrial laboratory that still does significant pure research is Microsoft Research. It enjoys an annual funding of about $7 billion, a level that can be provided by only a monopoly.
In Japan, the only industrial laboratory that does signficant pure research is NTT Laboratory.
The management of Mitsubishi Electric Corporation acted appropriately in shutting down the pure-research arm of Mitsubishi Electric Research Laboratory (MERL). Although MERL is part of the huge Mitsubishi conglomerate, it is not a monopoly in any industry and cannot afford pure research.
Never heard of them.
When old groups of individuals breaks apart, it shows the strength of the individual in society. Too often we think of "big bad corporations" who "control" their employees -- but in reality the employee is always allowed to leave. In some situations, employees make a bad decision and sign away their rights to compete, which only shows that the individual does not feel the risk of considering competing in a market is greater than giving up that ability in the future.
Everyone has talents of value to others. Sometimes that talent is not something you might want to do (for example, prostitution or sewing clothes), but keeping your individuality is key to building a better future for yourself and your family/household. When your employer doesn't offer you compensation (not always financial!) compared to the labor you expend, you can always try to go elsewhere. If someone else offers you better compensation (again, not always financial), you know the market works.
For me, I've never had a "real job" because I've always put my personal value above what any one employer could offer. I saw no reason to lock myself in, even if the financial reward might have been better. I've passed on jobs worth 3 times what I made in a year on my own because I kept my own individuality, and worked hard to build my value so I could exceed what the employer offered me. During that time, though, I was also free to pursue tasks and ideas that I would be locked out of if I had a salaried position.
This goes to show everyone that they should always look at what they're getting versus what they're giving -- and always consider what the market needs that there is a limited supply of. Point yourself in that direction, and grow beyond just being another W4. It can be done.
"They were the first group to develop the Diamond Touch table, an early precursor to Microsoft's Surface Computing."
MS 'Surface' table has nothing to do with touch. Below the glass 'surface' are five cameras - the device is simply a motion detector wired to a PC.
I am sure that Mitsubishi Bank, Mitsubishi Agricultural Machinery, Mitsubishi Estate Co., Mitsubishi Plastics, Mitsubishi Electric, &ct. would disagree with your "Core Competencies" analysis. The core product in the Mitsubishi brand is not cars or electronics. Their core product has always been venture research. Be the first and best in new fields. If anything, I am surprised they don't have a Mitsubishi Pharmaceutical yet.
MERL published a lot of papers, but that doesn't mean that they were doing research what was actually useful for anybody. The state of academic computer science is actually rather depressing, with lots of stuff being reinvented, meaningless variations of known techniques being published, and faulty mathematics being widely used.
The Diamond Touch table itself is symptomatic, since MERL didn't invent the concept either, and since it seems pretty clear that such a device does not have a significant market at this point.
Moreover, they seem to be doing a lot of deep computer work for a company that is little known today for computers, but rather their automotive and consumer electronics divisions.
There is a huge problem in the automotive embedded software industry having to do with reliability and productivity. I think a streamlined Merl has an opportunity to do extremely well in this area if they put their minds to it. I understand Mitsubishi is a member of JASPAR, the Japanese consortium funded by the likes of Toyota, Nissan and Honda. They recently announced the funding of a new automotive OS. Merl should focus on this more than anything else, IMO. Any breakthrough in this area is bound to spill over into other areas of computing and bring lots of profit with it.
Oh, just off the top of my head, uses for the first one: computer vision; second one, fluid sensors. And those are just the mundane ones I can pull outta my ass. The third one, UI, but I don't know where. But that's the point of research labs - they're supposed to come up with new stuff that can bring the company to new and more profitable directions, which is where the money's at, because old tech is commodity/low margin.
= l&q=l&c=msft
Whether the company applies all that neat-o stuff and makes money on it is not a fault of the lab, but a fault of management. Case in point: Xerox/PARC. "We're a document company, not a computer company" which meant that _everyone else made a buck but them_. No vision. None. Total Fail.
"Publically held corporations exist to make stockholders money, not to do research "because it's cool." Period."
Yah, like Microsoft, right?
Cool company versus Microsoft:
http://finance.yahoo.com/q/bc?s=AAPL&t=5y&l=off&z
--
BMO
Whether you are being cynical or not, science is not getting any help from your likes. Next time you read a scientific article, think of those words: WE ARE WEARY OF YOU KIND. WE DON'T LIKE YOU. GET LOST. AND FEEL FREE TO CUT OUR FUNDINGS --- YOU HAD ANYWAY.
Japan is different than America - there, even though they share the same name, the affiliation is much looser than you might think. Thus the core product of the Mitsubishi brand is Mitsubishi brand itself - and each of the affiliated companies has it's *own* core competency.
> according to Marks, who spoke with Xconomy at length Monday night.
So, TFA was written on the basis of a lengthy evening tirade by a disgruntled former employee. To their credit, they also interviewed the current CEO, who presented the alternate point of view -- that there was a gradual reorg in pursuit of a better ROI for the parent company.
I fail to see how this equates to "Mitsubishi breaks up MERL." MERL continues to exist. In fact it's a disservice to the newly hired researchers to assume that the "new" MERL can only be a shadow of its former self. Sometimes new people and fresh ideas are a good thing.
There is also a snide element of bias in the article against the "Japanese-style management" -- assumed to be something so horrible that researchers need to be "shielded" from it. While frustrating for some folks who can't bridge the cultural gap, this maligned "Japanese management style" is the same one that brought us innovations we take for granted today, in areas like automotive quality, 3G/4G cellular, the Wii user interface, CCDs and LCDs, and of course hentai anime.
In any case the news itself is interesting, but I'm not sure there is any need to portray it as the end of the world or something...
the golden era for applied AI research has gone for a long time. i don't believe there would be any new fundamental breakthrough (like logic-based reasoning and connectivism) solely based on math. if there is, those who really go for it would come up with it independently. it's not a teamwork-style research.
Back in the late '90s, I had a friend who worked for MERL. Can't even remember who it was, or how I knew him, now. Anyway, he was involved with the "Artificial Retina Skunkworks" there (Google Cache) and I had a Nec Versa 2000C running Linux (Red Had 6-ish).
:)
I forget whether I offered to play with alpha-test hardware, or he offered to let me, but anyway I wound up with a little circuit board with an A.R. chip on it, a 9V battery connector wired in, and a DB-9 serial port. Cabled it to my laptop, used gphoto (0.3ish maybe) to grab images with it. It wasn't high-res, and it was monochromatic, but it was kinda neat.
If I recall, the technology was to be used in a camera for the GameBoy or something like that.
Village idiot in some extremely smart villages.
Labs are out of date. Press releases that pump up the stock price before the next shareholders meeting are what really matters. Science takes too long, and gets bogged down in details. So Ladies and Gentlemen, let me welcome you to the wonderful new world of 'Faith-Based Research'. It's worked for climatology and geopolitics. Imagine what it can do for software.
Let me give you an example: Debugging. Ugh! It's rigorous and plodding. But with Faith-based Programming, close your eyes, concentrate and when you open your eyes, tell everyone it's gone. You will believe it, and so will they. If some pesky user tells you its still there, frown, tell everyone they're a trouble maker, meet with their boss and say they're undermining morale and suggest they're transferred or sacked. Beautiful!
Even though the US forceably broke up the Zaibutsu after the war, the affiliations are still strong (Keiretsu). The reason Mitsubishi Bank (now Bank of Tokyo-Mitsubishi after they merged with Tokyo Citybank) is one of the worlds largest banks is that all the Mitsubishi companies still do all their banking there.
They do.
Joe is a very impressive man--he can spot talent a mile away. The people gathered together at the offices in Cambridge were extremely good at what they do. I was lucky enough to have Micheal Mitzenmacher (an inventor of digital "fountains") as a co-author on a MERL research paper. Unfortunately, one of the results of the regime change seems to be the research paper store in Cambridge seems to have gone offline, so no link to that paper exists anymore.
The prototype bar glass that would signal the server when it was nearing empty was actually on a special table directly across from my cubicle. And the foot long white paper mache VW bug which served as a 3D screen for driving videos (you'd have to see it to believe it) was in the photo studio under the stairs.
As to whether Joe made good Mitsubishi products or not, I think it was obvious that he did not. But he was committed to making the lab world class (in a block crowded with world-class labs) which doesn't mean designing the Lancer's windshield wipers. However, since it cost him his job, maybe doing best for the world rather than for one's own advancement isn't a winning strategy.
Good luck, Joe, and thanks.
Bell Labs 1127 breaking up was a bigger deal IMHO.
Hope you like your current product lines, 'cause you'll be stuck with them for a while. Corporate R and D cuts are great for eliminating product migration roadmaps.
"I'm an old-fashioned type of guy. I worship the Sun and Moon as gods. And fear them."
The reason Mitsubishi Bank (now Bank of Tokyo-Mitsubishi after they merged with Tokyo Citybank) is one of the worlds largest banks is that all the Mitsubishi companies still do all their banking there.
Note that it's now "Mitsubishi-Tokyo UFJ Bank" because a couple of years ago The Bank of Tokyo-Mitsubishi merged with UFJ bank, which was itself the result of the merger of a bunch of smaller banks (including the quite large Sanwa bank).
It seems only a matter of time before there's only one bank in Japan, with a really long and severely hyphenated name... [Apparently "MUFJ" (as they style themselves) is already the largest bank in the world!]
We live, as we dream -- alone....
Yah, like Microsoft, right?
Cool company versus Microsoft:
I'm not sure what your point is. Microsoft does far, far more "because it's cool" basic research than Apple does, while Apple makes money for their stockholders. And I'm saying that as an Apple fan (and stockholder).
What I'm listening to now on Pandora...
depends on who they hire to replace the detectors, doesn't it?
Sounds like an aside from a 80's sci-fi novel.
Added Pressly: "Oh, and by the way, milk is nothing but liquid meat."
Come on, table computers were already production in 1986!! http://memory-alpha.org/en/wiki/Image:Galaxy_engin eering1.jpg And yes, as I understand it they were running Linux.
Which is the largest bank in the world depends a lot on how you measure size, but MUFJ doesn't do any better than 5th by any measure. The top 10 used to be dominated by Japanese banks until the property bubble in Tokyo burst in the early 1990's followed by the rest of the economy. They went from 8 Japanese banks in the top 10 by tier one capital in the 1980's (I used to have an encyclopedia that listed them, I think Mitsubishi was 2nd behind another Japanese bank, and Tokyo Citybank was also in there) to just 3. By other measures, they do less well, with at most one featuring in the top 10 now.
Nothing stays in its "original glory", otherwise we couldn't use those words now to describe it. People like to have a sense of nostalgia and this is no different. We tend to remember the good, not the bad. Right now there are other labs making history except no one knows or thinks of them fondly because nothing has fallen apart yet and there's nothing to reflect upon. It's all part of the technology ecosystem: produce great things, then break up the group and spread the love around, and mix it up a bit to produce new stuff the lab would have never considered, now that it's becoming a dinosaur. If every team stuck together companies would be full of satisfied employees not producing much except memories of the good old days.
So Mitsubishi goes the way of the once-great Bell Labs. Sad.
I guess it's the time machine version of Linux as Linus posted the first version of Linux in 1991.
That is very true. Unfortunately that is not what they do with their fantastic resources. Rather than try make things better by putting their efforts into improved products, they seem to put their efforts into gaining marketshare through destructive behaviour. Look at Zune and Vista. Look at WinCE.
Imagine how cool computing could be if MS instead put their efforts into competing by making better products.
Engineering is the art of compromise.