The Idea Factories

Bell Labs was a fount of important inventions, but it was not alone. Other industrial labs did too: IBM Watson, IBM Zurich, IBM Almaden, Xerox Palo Alto, Hughes Malibu ( with parquet flooring and a view of the Pacific, I might add), etc. All the way back to Menlo Park. But they’re largely gone.

We could resurrect the industrial labs. Should we?

This entry was posted in Uncategorized. Bookmark the permalink.

54 Responses to The Idea Factories

  1. sflicht says:

    Aren’t Watson and Zurich still going strong? Google X, Yahoo Research, Microsoft Research all exist, and seem to do stuff. My employer, a hedge fund, seems to find it much easier to recruit from Watson.

    On the other hand, the zeitgeist suggests that Big Tech is not exactly printing money with its basic research divisions. So maybe they’re just learning lessons that IBM, Xerox, and Hughes learned long ago. Focused R&D efforts, in the current economy, are perhaps a better investment than blue sky research.

    And how would you propose to “bring back” industrial labs, even if we grant the dubious premise that they disappeared? The one concrete proposal along such lines I could imagine is using NSF grant money to fund scientists working at these labs. That’s not a terrible idea, perhaps. But working out the IP ramifications sounds hard. Do you want Uncle Sam to get a share of licensing revenue from any patents that result from NSF-funded research? Should open-access publication be required for such research?

    • gcochran9 says:

      If you don’t know that places like Bell Labs used to be important and now barely exist, you are a confused puppy. Look it up.

      • sflicht says:

        I just don’t think your statement is consistent with the data. See here, for example:

        • sflicht says:

          Obviously that chart would be better if it were inflation-adjusted. But “industry share of basic research” is still an interesting metric and it hasn’t fallen off a cliff since Bell Labs stopped being important.

          Is your point just that industry is organizing its basic research differently? Possibly it’s true that the average size of an industrial basic research center has decreased.

          • benespen says:

            I don’t think corporate R&D expenditures are the right metric. I do things that fall under that number, but the kind of work I do is nothing like what Bell Labs used to do. Something about the way the work is done, and why, is very different, no matter how much money we throw at it.

            • barnyardboss says:

              “the way, the why” is indeed very different: predetermined outcomes to support corporate big pharma or big ag, or wall street climate scam derivatives. toss in some vibrant diversity affirmative action hires…. GIGO.

    • et.cetera says:

      No. Google X is a lot of good resources wasted on silly things. I’m not as personally familiar with Microsoft Research as I am with Google X, but I can’t imagine it being in a much better state.

      • sflicht says:

        Sounds like you’re an insider. What’s silly and why? I admit some of the projects seem dumb to me but what do I know?

        • et.cetera says:

          I can’t talk about stuff that hasn’t been made public. The main problem is X’s culture. Too much is bet on serendipity. It’s a cauldron for pet projects, stuff that is “fun” to work at but largely useless, so a lot of good, promising, interesting projects end up sidelined because they’re not as “fun” to work at.

          This “do what you like” approach works for the rest of Google because most of what the rest of Google is doing is boring. Here it’s a source of endless distractions.

          • sflicht says:

            Do you have any insight as to how Google X might be worse or better than predecessors likes Bell Labs?

            • et.cetera says:

              Bell Labs had the advantage of more, let’s call it “directedness”. The vagaries of a destination. Organisational issues aside, and I imagine our host will disagree with me on this one, X is in a worse position from the get go because a lot of ground has been tread. Bell Labs was in the right place at the right time and its big strides came in tandem with big strides in fundamental research in other places before and after WW2. I don’t think you’ll see a Nobel prize come out of X anytime soon. It’s hard to tinker in the dark and too easy to fool yourself that this or that gimmick will help improve the lives of millions.

              That’s why I can’t really give an answer to the opening question (should we resurrect the industrial labs). I don’t know that it’d make a difference either way.

      • nukeya says:

        I agree about Google being off in the weeds and needing direction big time. Too much fun and smugness of smart people who went to marginally better college than the ones working on the cash cow.

        However, that said, there’s a lot of VERY solid research work in silicon valley on deep learning and artificial intelligence now. NVIDIA, Facebook, Google, Numenta (Palm’s old Jeff Hawkins), and even parts of IBM Almaden are in on it. I think Tesla and SpaceX are doing some very good applied stuff. and there’s Steve Perlman’s group at Artemis doing some very interesting DSP work.

        • Scott Locklin says:

          Have you actually LOOKED at any of this research? Yes, the research is happening; that doesn’t mean it is any good. It just means the research is happening. Sort of like the research we did in the 70s on disco space colonies, and the 90s on “nanotech.” Maybe not that bad, but kinda close considering the hype.

      • Glengarry says:

        Here’s an interesting article on X:

        Sounds ripe for the chopping block once the good times are over.

    • CharlieK says:

      Before the AT&T/Bell System divestiture and the introduction of competition in U.S. telecommunications, Bell Labs had the immense advantage of its parent company’s government-sanctioned monopoly. I used to visit Bell Labs from time to time when I was an outside lawyer for the company, and it was clear that cost was simply not a concern. They hired brilliant people just because they were brilliant, and let them work on projects that might never flow to the bottom line. Easy to do when the company could go to the regulators every so often and ask to have its rates set at a level that would recover all costs plus a return on investment; and the regulators rarely inquired into the reasonableness of the costs incurred. No company facing competition, including today’s AT&T, can afford to be so profligate.

  2. JoachimStrobel says:

    And people flew to the moon during those times. This is not about industry promoting innovation, it is about the lack of it. During those time there were 4 billion people on earth, now there are 7. And their continental distribution changed dramatically.

  3. Gauss says:

    Not sure if they deserve to come back but it sure was fun working there. Not sure that Microsoft Research is quite in the same league. Yahoo Research?

  4. Ron Pavellas says:

    Ai and DNA-bioengineering labs. These seem prominent now. Not that I am celebrating.

  5. Maciano says:

    Marianna Mazzucato wrote an interesting book on the topic of gov-funded innovation. She opened my eyes to the reality that VCs are very short-term oriented and tend to focus on sure-wins. Private sector funded innovation works great, after the fundamental innovation has already been done. So, yeah, Western governments should fund and aid scientists and inventors to work on long-shots in Bell Labs-type organisations. 9 out of 10 times it might lead nowhere, but the 10th time you stumble upon big things that create huge new economic sectors. Current (low) economic growth rates prove Mazzucato is on to something.

    We can try squeeze out more and more out of the oranges we have, but it’s better to create more oranges so we get more juice.

    • sflicht says:

      At some point I thought her perspective was interesting but subsequently I decided she was full of shit, Why was I wrong?

      • Maciano says:

        You “decided” she was full of shit. That’s where you’re wrong.

        Her ideas didn’t jive with my personal preference for libertarianism either, but logic is on her side. I think you’re blinded by your ideology.

    • Greying Wanderer says:

      “She opened my eyes to the reality that VCs are very short-term oriented and tend to focus on sure-wins. Private sector funded innovation works great, after the fundamental innovation has already been done.”

      I think logic dictates there must be some truth in this.

      If you imagine all possible innovations some will be potentially rapidly profitable and can be funded privately but there must be some innovations that requires sub-components to be figured out first before they can become privately viable and if those sub-components are not potentially profitable enough to be researched then the main innovation will never happen.

      So the key for a govt. lab would be not trying to pick commercial winners – as private funding can do that – but doing the foundational work on pre-commercial components that makes them commercial.


      May not be a good example but i’ll use it anyway as i want to see it – biodomes.

      It seems to me there could be a lot of money in climate controlled biodomes in extreme environments: Antarctic, Siberia, Canada, Alaska etc plus deserts, prairies, underwater seaweed farms, moons etc, both as habitation and farming – once all the basics of manufacture and operation are figured out.

      That to me seems like the sort of area where commercial profit is too far away for much commercial investment and govt funding could get the basics done first.

      (Putin if you’re listening: biodomes, Siberia, hint hint.)

      • Greying Wanderer says:

        deserts too
        – shaded glass to reduce sun during the day
        – insulated to keep it warm at night
        – maximum water recycling
        – high tech desert farming


        just rambling now

        leapfrog off green arguments but turn them space age rather than stone age:
        – not enough food!
        grow food in Alaska and the Mojave biodomes
        – still not enough food!
        orbital farms
        – not enough metal!
        asteroid mining
        – not enough water!
        recycle it in climate controlled biodomes

    • nukeya says:

      VC’s are short sighted. Heck, I am too… I’m just trying to make a big score in my day job so I can pay off the mortgage.

      It’s the high-net-worth individuals who still have the drive and passion to use their money to make it happen – like Steve Perlman, Elon Musk, Jeff Hawkins that are the ones really pushing the limits with a long-term view.

  6. John Smith says:

    We should. Separating hard sciences from Cultural Marxist milieu of the university can only be a benefit to society.

  7. alex2 says:

    There are a bunch of big new industrial labs that do research (in the sense that researchers publish and aren’t force to work on company products) but they’re mostly focused on Machine Learning and neural networks more specifically.

    Facebook AI Research (FAIR)
    Google Brain (though some of them work on products)
    MSR (older and not as NN focused)

    My impression is that industrial research labs go in cycles. The companies start them when there’s rapid progress and they can only stay competitive in engineering if they also fund research. Even Apple announced that their new research group is going to publish papers. But when progress slows down they just fund engineering groups.

  8. albatross says:

    Bell Labs was a prestige project supported by the profits from AT&T’s monopoly, and started being downsized away a few years after the monopoly was abolished. I think you might say something similar wrt IBM and Xerox PARC, but I’m less sure of that. Other than Google, not many companies have some aoparently endless stream of wealth that’s more than they can spend, with which they can build this kind of prestige lab.

    What government policies would lead to more attempts to build another Bell Labs? Tax breaks for Nobel prizes?

    • Ursiform says:

      Xerox failed to take advantage of most of what PARC developed. IBM wasn’t as bad, but also missed opportunities. Not much point in funding a research lab if your corporate culture resists the innovations they develop. They may make for good public services, but that doesn’t make them good investments.

      • benespen says:

        This is exactly the problem. Bell Labs produced an amazing amount of good work, both scientific and engineering, but also AT&T managed to use their market dominance to suppress some ideas as well. In my opinion, this was probably a good tradeoff, but the things that made Bells Labs desirable for AT&T no longer pertain.

        Some combination of public-spiritedness and monopoly made Bell Labs work, and the lack of these things make most current corporate research departments markedly less effective in producing things that are not just good for the company, but good for everyone.

      • Peter Lund says:

        Xerox still made out like bandits on the laserprinter alone. PARC more than paid for itself, even if most such labs don’t.

  9. RCB says:

    So, where do the Claude Shannons of the world work now?

  10. engleberg says:

    ‘We could resurrect the industrial labs. Should we?’

    Yes. And we should dig up Jimmy Carter’s judge and hang the corpse.

  11. AllenM says:

    Idea factories are simply supported institutions with some common goals. Ideally, it makes sense to outsource this to universities to some extent.

    Now, back to breeding smarter kids- here is the drawback:

    Click to access 073452.full.pdf

    Key takeaway: ”
    Theories of ASD are numerous. According to the hyper-systemizing theory[31; 32],
    people with ASD have an unusually strong drive to systemize. A comprehensive
    hypothesis of ASD, taking intelligence and nearly all aspects of ASD into account, has
    emphasized the role of an optimum level of a suppressive force of innate traits[12].
    It has been reported that similar phenotypes and genetic parameters influence
    preferential mating[33; 34; 35].
    Consanguineous marriages appear to increase the
    prevalence of ASDs[36; 37; 38]. The preval
    ence of autism might increase by 1.5-fold
    after 1 generation of assortative mating (≥
    2.4-fold in the long term) depending on
    several assumptions[39]
    . Common genetic variants are individually of little effect but
    may be a major source of risk for autism[40]. Preferential mating may bring about
    additive genetic influences in concentrating inherited ASD susceptibility[41]. These
    observations suggest a potential role for combination of common variants in ASD.

    In other words, too much assortative mating may significantly raise levels of autism- of course, the results from our long term runs of Amish and now LDS should show the same problems- in short, smart people should seek mates that do not share a significant amount of genetic background, because of the tendency to express ASD in their children.

    Utterly fascinating.

    See who paid for this study:

    This work was supported by the National
    Natural Science Foundation of China grant
    81171880 and the National Basic Research Program of China grant 2011CB51001

    Why we are now not getting ahead in basic research- we need Chinese researchers to do basic science on the European genome.


    • gwern says:

      The autism common variants only correlate r_g=0.4 or so with intelligence (LD Hub estimate). This leaves a lot of intelligence gain without any autism risk increase. So you can avoid increasing autism risk too much by applying two polygenic scores and taking the best of a weighted sum. (And you can do this for indefinitely many correlated traits, and should do this instead of selecting only on one trait.) And autism currently isn’t that common, so an increase in autism risk isn’t that expensive. Remember, in a liability threshold model like for autism, the expected values are nonlinear due to the thin tails: a 1SD genetic risk increase doesn’t result in many additional diseases if the disease is initially very rare like 1%, but as population risk increases, the absolute numbers rise quickly – so embryo selection against schizophrenia isn’t that important in a couple with no family history of schizophrenia and thus a 1% probability of having a schizophrenic kid, but for a couple with schizophrenic parents or schizophrenia themselves, their kids may have 10-40% risk and then the embryo selection against schizophrenia becomes extremely valuable.

  12. j says:

    Big Pharma has “industrial laboratories” that do pure research. Each university has attached a business licensing company that sells its research products and patents. At least in Israel they do. The Weizmann Institute and the Technion make money partnering with the private sector. All that was unknown when Bell Lab thrived, today the same thing is done more businesslike.

  13. bob sykes says:

    A majority of the graduate students in our STEM programs at our major research universities are foreigners, mostly Chinese. Some stay, but the great majority go home because there are real opportunities in places like China. So we are transferring our advanced science and technology to China. Our manufacturing facilities themselves, of course, are gone. If anyone will ever develop industrial research parks again, I’d bet on the Chinese.

    You might note that right now China’s industrial base is larger than ours, and it is far more comprehensive: socks to supercomputers. With regard to the latter, China has more supercomputers than any other country or trading bloc in the world (several more than us), and it has the two fastest supercomputers, both home grown and built from scratch.

    • MawBTS says:

      With regard to the latter, China has more supercomputers than any other country or trading bloc in the world (several more than us), and it has the two fastest supercomputers, both home grown and built from scratch.

      Yeah, China banned US-made chips from their supercomputers in 2015 but since then they’ve only gotten better. Must be galling.

      Apparently the US Dept of Energy has commissioned some machines in the 150 to 200 petaflop range (TaihuLight is 93 PFLOPS) so maybe America will take the lead again.

  14. pdimov says:

    Isn’t, however, the right question “shouldn’t we resurrect the conditions that caused the industrial labs to appear?”

  15. I have two pals from those labs, who have different retrospectives on them. One worked in low-temperature ceramics and has patents which keep him fed for the rest of his life. He is glad to be away from people delusionally chasing Nobels, and now works in happy obscurity, joking about having finally perfected invisibility. The other was in soft solids rheology but I have no accurate idea what the hell he did. He still shakes his head and tears up that he has to “settle” for the jobs he has, as there just aren’t that many anymore. Of course both are in their early 60’s and represent a particular generation of researchers. I’m not sure you can get an objective view from old bastards, whose personal stories are too tied up in the answer.

  16. barnyardboss says:

    “an objective view from an old bastard”- well, maybe subjective- there is a entire generation of highly educated technicians passing into the night- people who could build the things the geniuses drew on paper, but didn’t know one end of a wrench from the other. dream away: without an 110-120 IQ technical class, the infrastructure continues to rot while we go the way of south africa- much less return to space.

  17. TG Cyrus says:

    Dr Cochrane, I am in deep mourning that I did not find your blog earlier. It is incredibly refreshing to find a rational empiricist in a sea of delusional zealots.

    Regarding the subject at hand, research must be one of the functions of government. Although I firmly believe that government must be as small and efficient as possible, there are a few critical domains which are theirs by necessity: law enforcement and research.

    Why research? Simply because it would not be accomplished any other way. Research is a lottery. The majority of gambles fail, but the precious few that succeed push the boundaries of human capability. Were it not for the government, the degree of scientific progress in the 20th century would be anemic. The Manhattan and Apollo projects alone constituted such density of scientific progress that modern endeavors of science seem miniscule in comparison.

    It could of course be argued that the state invests in science only when it is under existential threat, and history proves this assertion to be incontrovertible. The only alternative is to return the ancient tradition of patron backed science, where wealthy donors fund scientists of their choosing at their leisure. I believe this is the system that most closely resembles the future of science in the west, as shown by the involvement of prominent billionaires in funding their favorite pet projects. Of course, the downside here is that billionaires aren’t automatically qualified to select the best scientists nor the best venues of research, though it could be argued that neither are grant funding organizations staffed by scientists with potential conflicts of interest.

    The critical issue with science is that it has never been a profitable activity in the short term, and thus there will be very few profit conscious entities which engage in it freely. We must incentivize science, or see it wither.

  18. Bob says:

    Presumably you’d have to revive what Galbraith called the “technostructure” and the industrial state in his book The New Industrial State. In this arrangement, the modern corporations were run by large managerial bureaucracies and managers who didn’t own shares, rather than by the owners or shareholders. There is a good Adam Curtis documentary about how this model and managerial capitalism was dismantled in the 80s by the corporate raiders and shareholder capitalism:

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s