[jackpirate]

research of my own that'd be too risky for academia

I don't get this. The whole point of university research is that it is too risky (or benefits will be too far in the future) to make it profitable for companies to pursue. I don't see how anything could be too risky for the university, but not too risky for a business.

Edit: I'm not saying universities are perfect. I'm just saying that they still manage to do research that is too risky for industry. For example: finding the higgs boson. I don't see any examples of industry doing research that is "too risky for the university."

[Homunculiheaded]

That should be the case but it is not primarily for 2 rather related reasons: tenure & funding.

Everything you do as university research is towards the aim of getting publications to progress either you or your adviser towards tenure. The problem is that things that don't work are not usually publishable. This leads to 'fluffing' up results (read enough academic papers and you'll find some hilarity there) and avoiding anything that might not pan out.

Just as important is funding, even a small lab with a few computers and one or two grad students needs funding. And in CS that largely means DARPA or a handful of other government agencies. If your particular research interest involves eventually killing someone, you'll do fantastic. There are of course other sources of funding but they typically have much smaller wallets, especially the further you go on the 'for the good of humanity' scale.

I've seen countless times were grad students are doing something really interesting, but because it's not going to help anyone get tenure and not going to bring in any funding, these students are strongly encourage to 'get back on track'.

[_delirium]

If your particular research interest involves eventually killing someone, you'll do fantastic. There are of course other sources of funding but they typically have much smaller wallets, especially the further you go on the 'for the good of humanity' scale.

Unfortunately, being in the EU, I don't think changing this part fixes that many of the problems. EU grants are not about killing people, but typically about cooperation, building understanding between nations, reducing violence, integration of immigrants, that kind of thing. Goals I like more, in principle, than DARPA's. But the actual administration is if anything worse: 100-page proposals via hugely bureaucratic processes (NSF's are at least only 15 pages), unwieldy multi-country consortia, periodic mandatory status update meetings in Brussels, the works. If you can sell what you're doing within that framework and work with it, it's good, but it's still very much about being on the right track to impress the purse-string holders.

[xtracto]

But the actual administration is if anything worse: 100-page proposals via hugely bureaucratic processes (NSF's are at least only 15 pages), unwieldy multi-country consortia, periodic mandatory status update meetings in Brussels, the works.

Tell me about it. I was 7 years in "academia" in the EU and 4 of those years was working in a FP7 (Seventh Framework Programme) project. Man, the amount of administrative baggage we have to do is amazing. "Libre research" is a myth, between monthly deliverables, 6-month review, half-way Brussels project review, you could not make a lot of research.

And then you have all the administrative controls, I kid you not, I had to log what I did every day in two places (one in EU FP7 timesheets (in Excel), and another one in my institute's own software (a terrible java program)). Sure, the upside was that I got to travel a lot (the project had about 7 participant countries).

Now I returned to the industry, I am in a company from the Silicon Valey as a "simple" software engineer (even though I have a PhD) and I could not be happier. Moreover, after a couple of months in the position I have made several contacts which are bringing new opportunities.

I am happier at my current "fast paced" job now as I was while I was in academia doing papers just trying to publish papers for the sake of it (that is how I felt).

[tikhonj]

Don't some of the large software and hardware companies also support research at universities? I think that some of the research labs at my university get supported heavily by companies like Google, Microsoft, Intel and Qualcomm. However, I don't really know the specifics.

[Homunculiheaded]

They definitely do, I didn't mean to imply that DARPA is the only source of funding. Usually for most universities you can look at Large Company X, that operates locally and find them contributing to any department that does the kind of work they need. For example I believe Microsoft heavily funds the CS department at the University of Washington.

Of course this doesn't get you out of the trap of having to do research in the area of interest to whatever large company is funding you. You can do research that isn't indirectly getting people killed ;) but it usually has to be pretty clearly aligned with the business interests of companies funding the research

[tikhonj]

Heh, I guess being in the Bay Area really helps with getting a really wide range of companies to sponsor research :P. Looking through some of the bigger research labs, it seems most big companies represented at some point.

I was just curious because my impression is that CS is rather well funded, especially for having far lower expenses than anybody else in college of engineering.

[aswanson]

I don't get this. The whole point of university research is that it is too risky (or benefits will be too far in the future) to make it profitable for companies to pursue. I don't see how anything could be too risky for the university, but not too risky for a business.

I get it. I recall specifically posing a promising, if not clear-cut, research topic to my grad school advisor and his response "that's high risk".

Academia is more concerned with the volume of publications rather than impact and unique ideas. There are tremendous disincentives in the system for innovation.

[_delirium]

You can blame citation-counting and h-indexes for a decent amount of that latter part. If administrators who don't know enough about an area to individually judge the quality of someone's record are mainly using citations and h-index as proxies for quality, then the job of a junior researcher becomes to maximize those two numbers. There are various strategies for doing so, but one of the higher-probability ones is to churn out a lot of papers, and also to make them relatively similar to what other people in the field are already doing (so they'll cite them as related work in their next paper).

[neilc]

Academia is more concerned with the volume of publications rather than impact and unique ideas.

As a generalization, this is false. Certainly there is more emphasis on publication than is warranted, but most of the top places care much more about impact/reputation than about raw # of papers or citations. If you talk to people about how research labs hire newly-minted PhDs, how faculty are hired, or how tenure decisions are made, everything I've heard suggests that, at least for good places, they aren't just looking at the length of your pub list or the # of citations.

[amirmc]

"There are tremendous disincentives in the system for innovation"

I feel that's overstating it but even if not, is it worse than industry? I've found academia to be very accepting of different viewpoints and depending on the field, experimentation is encouraged.

In my viewpoint, if an advisor in academia calls something 'high-risk' it's within the context of the overall work. For example, in grad school, the aim is to make a contribution to the field and get a PhD. Several students I've known have followed 'promising' lines of work (after disregarding advice to contrary) and ended up with nothing.

[carterschonwald]

There's that issue too, and very true. (though what I'm currently up to is a project I started after I left my program)

[madhadron]

That's the point, but the reality of job, training, and funding structures today is that almost no one in academia can do anything that is not an incremental step on the research who has already managed to become politically established.

[jackpirate]

Right, but that's pretty much true of industry as well. I just don't see any examples of major scientific breakthroughs coming from industry that aren't equally coming from universities.

[mturmon]

But, returning to the particular case of the OP:

His essay seemed to say that the cautious incrementalism and turf-buildout that succeeds well with NSF (where most of the funding for academic ML research comes from), turned out to be less exciting than riding upon the Google juggernaut.

It's hard to argue with that. The applicability of ML technology has attracted companies with deep pockets, and really changed the landscape of research.

[jlarocco]

I think that's covered in the essay, actually.

He claims universities are increasingly judging professors by their ability to get funding. And, due to budget cuts, funding is only going to areas likely to pay off. "Too risky for academia" would be too far off the beaten path to get funding.

That said, I don't know if it's true or not.

[_delirium]

I agree with that, though that's still a somewhat different idea of "risk" than companies have. It's perfectly possible to do "risky" research in academia in the sense of something that may not pan out into practical applications in the next 50 years... if a funding body has made it a priority. If the NSF has decided that Foo Research is a priority and will be funded at $100m/yr for the next 5 years, then it is not risky for the professor to research in that area, but it might be risky research in the sense that it may never pan out to be worth that $500m NSF investment. In industry, that kind of research may never have happened.

What's increasingly hard to do in academia is to take risks individually: if you decide something is under-researched and should be a priority, rather than waiting for a funding agency to decide it's a priority, you'll have tough going.

[mturmon]

The hunt for the Higgs boson is going on at CERN, which is not a university. It's more along the lines of a national lab.

[zbyszek]

It's more along the lines of an international lab. Many, perhaps most, of the academics working there are still employed by their home institutions and are funded just as if they were working in a lab in their own university.

In fact, the same can be said for some of the academics working in national labs.

[mturmon]

National lab, international lab, whatever. Point is that CERN is still more like a large, government-supported lab than a university.

What's happening at CERN is not at all reflective of the university environment of the OP, which emphasizes small collaborations working on, say, million-dollar projects lasting 3-5 years. As we both know, CERN is nothing like this.

I commented because the idea of saying that the work at CERN is "university research", as was done in the comment above, is mistaken. CERN has no relation to the work done by the OP.

[mrj]

Business is perfectly willing to fund 10 ideas if one success will produce an outsized profit. Academia is not so kind.

[jackpirate]

Two enormous successes of academia are mapping the human genome and finding the higgs boson. Both were hugely speculative the time the projects were undertaken. If there's something comparable coming out of industry, I don't know about it.

[mbreese]

The human genome sequencing didn't really get going until they had to start competing with Celera. In the end, the HGP ended up adopting the same techniques to keep up.

So, that probably isn't the best example to use.

The LHC on the other hand is one example of what the original post was worried about - the trend towards more consolidation, towards funding bigger projects (with fewer people), at the expense of many smaller labs.

[tensor]

This is very misleading. Celera used techniques born from academia to sequence the genome quickly. While it did prove that the method was both feasible and preferable, it did not cause the genome to "really get going." The genome was nearly finished already by the time Celera got close to completion.

The public draft was also of much higher quality then Celera's, and as hindsight shows, also born of better scientific technique. Celera secretly used only one person (very nearly) to sequence from, rather than a collection.

If Celera never existed, the human genome would still have been completed in a similar timeframe and shotgun sequencing would still have been adopted. If nothing else, academics are just as competitive as anyone else and thus some lab would have loved to show the feasibility of shotgun sequencing, albiet on a different species.

Disclosure: I worked on the public draft and also had access to the Celera draft right around the time that the genome was "completed". Our lab was using and comparing both the public draft and the Celera draft at the time. We were even fusing the two to get the best possible "up to the minute" draft for a specific chromosome.

[scarmig]

The Higgs boson also wasn't super risky an undertaking. It was expected, and the project was a multi-billion dollar affair that was likely to come up with something. A somewhat ungenerous view is that it was just glorified taxonomy.

When you think about the sums involved, we could have done a whole lot of smaller, riskier projects, which may or may not have panned out but had a great deal more potential. Which I guess is your point.

[neutronicus]

It's not like looking for the Higgs is the only thing people have done at the LHC. For instance, there's another big experiment going on investigating Quark-Gluon plasma, and the OPERA experiment was using LHC neutrinos.

The LHC is a big, expensive piece of infrastructure that groups have to share, but there are multiple groups.

[yen223]

Could the smaller projects have the expertise and resources to discover something like the Higgs-Boson?

Centralization of resources is not necessarily a bad thing.

[SoftwareMaven]

The more important question is could multiple smaller labs have come up with something more valuable (to science or society) than weighing the Higgs Boson (the discovery was expected, the weight was an unknown)?

[me2i81]

You can like Celera or hate it, but you can't just pretend that it didn't exist...

[pmb]

It was, however, founded by a guy who left the human genome project and took his hard-won expertise elsewhere.

[angersock]

i mean all weve accomplished is connecting over a billion people and letting you see the person you want to talk to in real time on a device smaller than a hotpocket

yeah i guess industry isnt really accomplishing much these days

(How are your examples actually useful?)

[mbreese]

And most of that was driven by academic research 10-50 years ago. The fear is that we aren't currently funding the academic research that will drive technology in the next 20-50 years.

[angersock]

If you take that timeframe, fair point.

[sthatipamala]

Sibling commenter explained it fine, but to elaborate... your example relies on the existence of wireless networks, LCD technology, assorted networking protocols, cryptography (for secure logins, etc), and video compression algorithms. All of those have been advanced in some part by academic research.

[mechanical_fish]

I believe the quote refers specifically to the speaker's research, not all the research on earth.

[carterschonwald]

Yup

[mfitzerald]

With business, if you don't innovate, you die. University research is usually grant-driven. Grants are the result of filling in the right answers and telling a committee what they'd like to hear. There's much more urgency to be constantly innovating in business.

[yen223]

With business, your innovation absolutely has to turn a profit. This means that important research with less potential profit will die out, e.g. malaria and tuberculosis treatments.

It is for this reason that I believe there is still a place for academia in our society.

[alextp]

Regarding finding the higgs boson, regardless of whether you agree that it's what they're actually doing, the main people actually building the first quantum computer are D-Wave, a private company.

[gcr]

citation please? First, what does quantum computers have to do with finding the higgs boson? Second, is D-wave not standing on the shoulders of decades of academic research in designing their computer?

[lmm]

D-Wave are making more exaggerated press release claims. Universities have better working quantum computers (i.e. any at all)

[neutronicus]

They've got some serious-business Quantum Computing experiments at NIST, as well.