The death of Intel Labs and what it means for industrial research | Hacker News Reader

The death of Intel Labs and what it means for industrial research

matt-welsh.blogspot.com

  These companies can afford to set up big labs with lots of PhDs and 
  pay them to do whatever the hell they want with little accountability

As an anonymous commenter has already pointed out, with the possible exception of some parts of MSR, there aren't any industrial labs where you can "do whatever the hell you want". I was offered a position at IBM research and they were very clear from the outset that they had a specific set of projects they wanted me to work on. In fact, these projects had less to do with my research work at grad school and more to do with what I'd done in my previous life as a software developer. These sense I got from talking to the researchers at IBM was there is a clear expectation that IBM research would directly impact the company's bottomline. The group I was to be recruited into was helping solve some of the hard problems that the rest of the organization simply doesn't have the expertise to solve.

I dunno about other fields but in my area of work, which is computer architecture, MSR, Intel and IBM have all written important papers that have directly influenced the design of many current-generation microprocessors. So, I'm a bit hesitant to suggest that the labs have failed or that the model is broken. The model could do with some improvement for sure, but then again, what couldn't?

There is an interesting point-of-view in the comments:

"Microsoft is different because it is a monopoly-supported research lab. The old Bell Labs had a monopoly underneath it to support it: costs AT*T incurred to run it could be charged back to consumers. In this case, the monopoly is windows/office, but that will slowly fade. So my belief is MSR will eventually go down this road, it just may take 10 or 20 years. Some friends of mine there also believe this."

In the past years, I have seen different research works funded by Microsoft Research where is notoriously difficult to use other technologies (e.g. based on FLOSS) than the one distributed by Microsoft. It seems to be a limitation factor to innovation or research because you restrict yourself to an existing technology or product line. Is this still research?

MSR probably paid for itself with ClearType alone.

But that is a foolish comment; AT&T really was a monopoly, it had the government enforcing that on its behalf. It was effectively a branch of the civil service and its lab an academic department. Windows/Office cash-cow is really no different from IBM's mainframe cash-cow. They're in a dominant market position, sure, they're cash-generative, sure, but if they don't do research, their position becomes ever more uncertain.

>AT&T really was a monopoly, it had the government enforcing that on its behalf.

And weren't its tariffs generally set to allow it a certain profit margin? In that sort of setup, the decision to fund a big, expensive research lab is a no-brainer: you're guaranteed not to lose money in the short run, and the lab will probably make you money in the long run.

On your FLOSS comment: the agreement for a MSR PhD grant is very clear about "absolutely no GPL ever". But that's it. It's not hard to find MSR research into, say, OpenSSL running on Linux.

(Similarly, they demand a free license to any patent that comes out of your work, but the patent is still yours/your university's. I've heard rumours of worse contracts, but this one was just about being able to use the results of your work.)

I do not agree with the author's conclusion that the model of industrial research (MS,HP IBM etc) is broken. We need places in industry to do basic reserach, if everyone was focuses on engineering or product based research (Apple, Google) it is not good for the over all ecosystem. Many inventions are made by pure serendipity we need to give the brilliant minds the freedom to pursue their research based on curiosity and interest.

We do need to pursue it, but a short-term oriented market based system just can't do it. There is too much financial risk in the kind of research that needs to be done. And if you do ever manage to produce anything, everyone else can just copy it. So the best strategy in the market will always be to not spend anything on research and just steal from those who do.

Is there a solution to this multiplayer prisoner's dilemma/coordination problem besides government-run research labs? I've never heard one.

Patents are the solution our society seems to have chosen. I don't know whether they're "the" solution.

Yes and no. HP, for example, invests a lot of R&D money in the core businesses and not everyone can just copy what we do and even though quarterly reports rule there are 5 year plans for things other than off-the-shelf consumer products. On the other hand, the D of R&D is heavily emphasized and HP Labs has been struggling to be relevant to the business units as long as I can remember. The question for blue chips is whether it's cheaper to fund your own research or wait until something is mature to the point where you can buy or license it. Much as it dismays me sometimes, it really isn't such an unreasonable strategy for the average stockholder who owns us.

> but a short-term oriented market based system just can't do it

How can you know for sure? There is no such thing as a free market system in the United States. The incentive for private research is very low when all companies have to do is wait for subsidized research coming out of universities.

Title: "The death of Intel Labs and what it means for industrial research"

Line 3/4: "I should be clear that not all of Intel Research is closing down -- just the [3] lablets."

...no comment

Intel Labs is a part of Intel Research consisting of the three lablets. It is closing down.

Check http://www.intel-research.net/

So they're redirecting the money towards universities? That's great -- in principle, at least. Research is their core business.

However, in the current financial/academic climate, I suspect that universities will simply withdraw internal funding from these departments (because, well, the CS guys got their money now) which would result in a net loss.

A university that has a reputation for doing this will attract less funding.

This isn't a new phenomenon. Most of the money that comes to successful departments comes from research funding, usually from the big funding agencies, and the risk is more that the university admin lavishes attention and resources on the most visibly successful departments, and so sacrificing breadth and the strength of their interdisciplinary work. Look at the recent Kings College London fiasco for a recent example.

The article is part of an old problem: How to connect 'research' with 'business'.

I respond with:

(I) Summary.

(II) Details on Reasons for Failure.

(III) Examples of Solutions.

(IV) Opportunity.

below.

(I) Summary.

Quite generally, there are many ways to make a mess and fail, and this is true also in connecting research with business and has a very long history.

There is also a long history of success connecting research with national security, public health, and business.

Net, the main issue is, are results of research really wanted outside of research? The usual answer is, "Not only no but hell no!", and that's the main reason for failure.

(II) Details on Reasons for Failure.

One level deeper, the main reason people outside of research don't want research is that they feel threatened: They would rather just keep doing things the same old way that has been doing well for them (people not doing well can't afford research!) and not risk it by being 'upstaged' or 'disrupted' by research.

Also a researcher with good access inside a successful business organization can be seen as a 'loose cannon on the deck' that might understand the 'secrets' of the success of the business, use research to find a better way, and start a competing business.

The usual situation in business is: (1) Organize like at Ford 100 years ago where the supervisor knows more and the subordinates are just to apply labor to the ideas, direction, and work of the supervisor, (2) pay everyone relatively little, thus, making sure no one feels free to take any risks at all, (3) evaluate everyone on very narrow criteria of getting the present work done with severe consequences for any deviations and, thus, for anything new, ensure no upside and a lot of downside, (4) for anything new inside the organization, kill it off ASAP, (5) for anything very new or advanced outside the organization, f'get about it.

So, net, the only one in the organization with any freedom to do anything new is the CEO, and generally he is too busy just running the existing business and not technical enough to do or supervise anything like 'research'.

Inside a business research lab, usually the goals are at best a mess. So, the path to the business bottom line is not wanted or not clear, but the usual academic criteria do not apply either.

So, without good criteria, the organization is vulnerable to the long, standard list of organizational dysfunctionalities. E.g., the place can become a 'caste' system where the head guys defend themselves with 'cliques' based on incompetence, race, gender, ethnicity, etc., and all the non-manager "worker bees" are expected to keep quiet and out of sight and maybe just publish papers or patents.

So, usually big business organizations just will NOT 'innovate'. This is a problem, but its 'flip side' is an opportunity central to 'Hacker News'!

(III) Examples of Solutions.

Again, one of the crucial points is for the people in the business actually to want the results of research, and at times there have been solutions at AT&T Bell Labs, DoD, NIH, CDC, etc.

One solution has been illustrated at Renaissance Technologies run by Jim Simons: He is a good mathematician (Chern-Simons), long Chair of the math department at SUNY Stony Brook, and at Renaissance sometimes paid himself $2 billion or so a year. Once the Brookhaven National Lab people wanted to collide some gold atoms to create a 'quark plasma', couldn't get the money they needed, so Simons wrote them a check, maybe $20 million or so. So, he wanted the results of research, had a reputation for hiring mathematical physicists from Russia, and was not threatened.

Another example has been Westvaco Paper: They specialized in 'specialty papers', e.g., milk cartons and table tops. Basically they have regarded themselves as an 'applied chemical engineering' company.

They had a research lab between DC and Baltimore: Each six months the lab submitted its budget to the NY HQ, and HQ always wanted to increase the budget, and Research always said, "No.".

Why? My old notes have:

(1) The research group worked only on problems of their own choice and did not accept problems assigned to them. Generally in research, problem selection is both important and difficult, and here the real experts did the problem selection.

(2) Projects started by the research group ran from about two months to about two years. Only about one project in ten led to an implementation.

(3) When research believed they had a project worthy of implementation, they would approach the General Manager (GM) of the appropriate operating group and make their proposal. The decision to implement was up to the GM. Here we see some the importance of having the business organization involved in the implementation.

(4) If an implementation was made, then the financial value was measured by careful auditing.

(5) For the first three years of implementation, half the financial improvement was credited to the operating group and the other half to research. After the first three years, all the improvement was credited to the operating group and none to research. Also, the GM's bonus could be affected by these results. It did help that most of the GM's were Ph.D.'s in chemical engineering.

(6) With these rules, the research group returned to the company, long run, about three dollars for each dollar in their budget.

So, that's one way to run an industrial research lab.

(IV) Opportunity.

For the 'Hacker News' community, the flip side of the sad, old situation is a new opportunity:

(1) Find a problem where a much better solution should lead to big bucks.

(2) Have the founder, CEO do some research to get a much better solution, good enough to be much better than anything else and advanced enough to be difficult to duplicate or equal and, thus, provide a good 'technological barrier to entry'.

Note: IMHO, for the relevant research for 'information technology' the key is not 'computer science' but selected, advanced topics in pure and applied math.

(3) Implement the solution in software and offer it to users via the Web or cloud.

(4) Buy a nice yacht, rinse, and repeat.

For the research part, the main training for research is the Ph.D. The difficulty of Ph.D. programs shows that learning how to do good research and doing it are not easy. Thus, hopefully, the founder, CEO doing the research has a good, relevant Ph.D. This way, no one in the organization is threatened by the research that helps start the organization.

Once again, the ways big organizations find to make messes can be corrected by entrepreneurship.

So, don't curse the big organizations. Instead, use their problems as an opportunity to be a successful entrepreneur and buy a yacht.

A famous recipe for rabbit stew starts off, "First catch a rabbit ...."

A recipe for applied research starts off, "First get an application ...."

The article has,

"These companies can afford to set up big labs with lots of PhDs and pay them to do whatever the hell they want with little accountability, but maybe this model is no longer sustainable."

Thanks a lot former Harvard prof: That's an insult to Ph.D. holders, cuts them off at the knees, and is a great recipe for Ph.D. unemployment: Or, just why will the lowest non-Ph.D. manager in a management chain want to bet their budget and part of their career on "PhDs and pay them to do whatever the hell they want with little accountability"?

Just what is it about wasting time you find so attractive?

If you want to sit in a small, closed, dark room and engage in intellectual self-abuse, first make, marry, and/or inherit some money and then go for it. For the rest of us who sacrificed a LOT and got a Ph.D., JUST to be, and ONLY to be, better qualified to contribute to both top and bottom lines of accounting income statements, please cut the insults.

Reply on news.ycombinator.com