The Two Cultures
In 1959, C.P. Snow gave a lecture entitled “The two cultures” – about, to quote the Wikipedia article:
Its thesis was that the breakdown of communication between the “two cultures” of modern society — the sciences and the humanities — was a major hindrance to solving the world’s problems.
Right now, I wonder if in the fields of electronics, computer science, and embedded systems, we see a breakdown of communication between our “two cultures” – industry and academia.
Given the state of the world economy and its impact on industry, I see a fair bit of “battening-down the hatches” among design groups. Thankfully, designs are continuing, but there is lots of attention to the nuts and bolts of design, without much time for more speculative research-oriented work. In many ways this is a continuation of the increasing shift towards pragmatic design that evolutions and consolidation in industry and consolidation has fostered the last few years. Another thing that has been scaled way back, at least by my observation, is the industrial participation in conferences, symposia and workshops, especially “scientific” (non-profit, sponsored by groups such as IEEE and ACM) ones, although commercial conferences have also been trimmed.
But when I look at my email inbox and the stream of announcements of conferences, symposia and workshops, I see in general no lessening in their number nor in the attractiveness of the locations (Greek Islands in summer continue to be very appealing although quite unlikely). And the programmes and organising and programme committees seem to be dominated more and more by academic research, with industrial participation in papers, presentations and attendance diminishing rapidly. The range of topics seems to become ever more esoteric and divorced from practical design interests and experiences. This is reinforced when I review papers for conferences and workshops that seem to pay attention only to the results of the research community and have little awareness of the state of the art in industrial practice.
Of course, there are many reasons why academic research seems to become more and more esoteric. There is a drive in academic research for novelty at all costs, no matter that industrial practice falls more behind. There is a bandwagon effect in academic research, where new ideas (ranging from NoCs to quantum dots) become popular due to their novelty and the research community can build a large edifice of results even if the real usage of such results in real designs is minimal. There is the need to obtain research funding, primarily driven by novelty – and a corollary lack of interest in the concrete problems faced by real designers, which are often considered as being “solved”, mundane and of little interest. There is research self-constrained by access to tools and mechanisms – which is one reason why I think there is an over-emphasis in academia on reconfigurable designs and computing – because FPGA-based systems are now much more available in the research community than access to leading edge SoC design.
But I also wonder whether the increasing distance between the communities is going to lead to a serious lack of innovation in industry, especially as it comes out of the current economic situation and will need new innovative ideas, both in a product sense and in design implementation. And the communications between the “two cultures” that might improve the situation seems to be becoming less and less likely as their interests and motivations diverge.
“What we’ve got here is a failure to communicate”
Paul Newman as “Cool Hand Luke”
Is this just me, or do others see this phenomenon as well? Are we communicating between the cultures as much as we used to? Are we talking as much as we should?
5 Comments »
RSS feed for comments on this post. TrackBack URI
Leave a comment
Line and paragraph breaks automatic, e-mail address never displayed, HTML allowed: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>
February 5th, 2009 @ 3:44 am
Great article. I have about 15 years experience in industry and a masters degree and was looking to continue further but see very few programs that target what I want. It seems in the USA, universities offer a Masters degree in engineering but only a handful go further and offer a Doctorate in Engineering. Instead, they all focus on a Phd which are almost allways described as preparation for those wishing a career in acadamia. To me it seems they are ignoring the large number of Phds who go into industry and who would perhaps benefit a Doctorate level education but with a focus advanced on technology AND on management. The requirement for novel research could be replaced by a project which shows a demonstrated mastery of a particular engineering domain. Texas A&M has a program close to this. Southern Methodist University also has one in Software Engineering. USC has the concept of an Engineers Degree which is close. But such programs are certainly stand-outs.
To me this propetuates the two cultures. Doctorate should not have to mean scientist.
February 13th, 2009 @ 8:52 pm
Grant,
I think you are absolutely correctly in pointing out the widening gap and disconnect between academia and industry. Personally I think one of the best examples of this is in an area of interest to us both – programming of multicore devices. It’s clear that this is a new and significant challenge but how have both camps reacted? Academia has generally reacted by buying a brand new and very large whiteboard and is busy inventing new languages and sophisticated automatic parallelizing compiler approaches. Industry is doing what engineers do – using what they have already built and what they have lying around them and making it work; not optimal I am sure but it gets products to market. This is of course the rationale behind the formation of the Multicore Association’s Multicore Programming Practices (MPP) working group. We are pulling together best common and sensible practices in use in industry today for use by the rest of industry who are still climbing the learning curve of parallel programming. The MPP team is a great group of companies and people – and even includes a select couple of practical academics! Maybe academia and industy will eventually come together but if they do I expect I will be resting comfortable in my retirement home. In the meantime I feel sure that academia – if it took a more pragmatic, evolutionary and pragmatic view – could add significant value to industry efforts and I hope they do.
David Stewart, CriticalBlue and Co-Chair MPP working group
February 15th, 2009 @ 12:40 pm
Grant,
Great article, I could not agree more. R&D is a poorly understood acronym.
Doing something more than D, a pure short-term product development, is far from easy in an industrial context (I was lucky enough to have this chance for 1 project).
On the other hand, an R completely disconnected from realistic concerns and a potential D is the preferred road on the academic side. What matters is that ideas explored in R look new and are able to attract funds.
By trying to bridge both sides, you end up in an uncomfortable seat where your R seems too pragmatic from an academic viewpoint (as I could often observe from paper reviews during my PhD) but not practical enough from an industrial perspective.
I am still trying to figure out how we could bring efficiently the two cultures closer, but have not found the answer yet.
Besides, a parallel “vertical” gap (the one you pointed out could be represented as “horizontal”) occurs between system- and hw designers, where the former come up with algorithms and solutions which are impossible to implement by the latter. The bad news here is that this situation is getting worse with outsourcing and globalization, whereas cross-disciplinary collaboration and understanding remain rare on the academic side.
February 17th, 2009 @ 12:44 pm
Good commentary Grant. Having spent quite a while working on both sides of the fence and also evaluating research, while working in the industry, I can agree with most of your points.
However, the biggest disconnect I’ve seen is one of scale. Research projects always seem too small to hit the really fundamental, yet mundane critical problems that are facing real industrial design teams.
Silly, trivial, solved problems like source control, or bug tracking or project management. Problems of complexity and scale that are never encountered in research environments because the problem sets are too simplified or abstracted to encounter the real issues that stall real world chips every day.
Innovation does happen in the small areas that get the research focus. Inside processors, or in reconfigurable technologies. These can be deployed as IP blocks within the big, complex chips that huge teams of people struggle with every day. But the issues of scale and communication and complexity management that seem to elude the grasp of most design & verification teams haven’t seen much improvement in the last decade or so, as far as I can see.
For such solved, mundane, uninteresting problems, people seem to be reinventing a broken wheel every year. It isn’t sexy. It probably doesn’t get grant money and it is hard for 5 person or 30 person research groups working with a high turnover of students, for a few years at a time, to ever even experience, never mind improve.
February 17th, 2009 @ 1:40 pm
Thanks for all the interesting comments. To pick up on Gordon’s point, when I was at BNR/Nortel (going back to more than 15 years ago), working in the “CAD” group, we were always interested in the notion of what I called a “Quasi-Realistic Design Example” (QRDE). This was something big enough to demonstrate the essential qualities of real designs, but small enough to be a reasonable test case for new design methods or tools or to pipe clean a new or improved design flow. In addition, if a QRDE was not so sensitive, it could potentially be shared with the academic community or with EDA tool suppliers. I have to admit that we knew what we were looking for in the QRDE, but very rarely found one! And designs have grown so complex these days, and the sensitivities around them so much greater (in the “old days”, a single ASIC was rarely anywhere near the whole of the “design”), that the idea of research groups in academia having access to realistic design examples and their properties (let alone whether these are seen as interesting), may be hard to achieve.