Missing the Point: Semiconductor Industry / Academia Critique
, 2011年01月27日
I thought for a few seconds, then looked him in the eye: “Because much of it is off target.”
My friend was taken aback for a second, but to his credit he wanted to know why I thought that.
I explained that I recently read an article in the March 2009 issue of the Journal of the American Society for Information Science and Technology called, “Design engineers and technical professionals at work: Observing information usage in the workplace,” and learned that published research was lowest on engineers’ list of sources of information. (The first sources were talking to colleagues and Internet searches.)
Because most academic researched is searchable online, I surmised that the issue was not the ease of availability of the academic research but rather its content.
Most academic research in the engineering disciplines focuses on easy-to-set-up problems, rather than more complex real-world problem statements. For example, in the SoC world you will find many research papers that use relatively simple topologies with homogeneous open source components.
Why is this bad? Let’s stick with my SoC example above: Although there may be production chips that mirror the experimental setup of some academic papers, the vast volume of embedded consumer electronics chips most of us deal with are very complex, with diverse topologies and heterogeneous IP blocks with varying latency, bandwidth and power requirements. Therefore, the answers learned through academic research solving simpler problems may not be applicable to the real-world problems seen today.
Why is this so? Upon further discussion I learned that academia and industry differ in two aspects: Timeframe and resources.
Academics generally want to solve really hard problems that will occur years down the road. What this means is that the answers they get today may have no practical use for years. Therefore an answer that today’s research is “off target” for today’s industry practitioners might be expected. And an academician that wants industry to be able to use today’s research might be more attracted to “applied research” rather than longer-term “pure research,” and there can be a bias in academia toward pure research.
Another reason is that it is simply easier to use readily available open-source materials and simpler, bounded problems in an experiment than it would be to create a more realistic (at least to today’s industry practitioners) experimental environment. Working with companies to do this type of research is extremely difficult because creating a realistic environment would require the academic team to:
- Work with leading companies to understand their technology needs and gaps;
- Obtain the same inputs and materials used by industry to create the test environment, and
- Obtain the same tools used by industry to measure and report the results in a manner consistent with industry practice.
It is obviously easier and faster to create and perform experiments, and write the resulting papers, using easily obtainable inputs and knowledge versus creating the relationships required to understand industry’s problems, obtain their support, garner financial resources to spend, and mimic their problem space environments. After all, there is only so much time in a semester or school year, and only so much money in a professor’s budget.
Who is responsible? We all are. Academia really is “publish or perish.” Professors and their students need to create a flow of papers and presentations to achieve tenure, maintain credibility and give Ph.D. students hard problems to work on to achieve their credentials. Governmental organizations offer research grants to institutions for projects, but sometimes the problems in the grants do not match up with what industry would like to understand. And industry often sees academia as a distraction and only wants to engage with them if it directly aids their technical and business efforts.
So what can be done? From what I learned in this discussion, I’d like to see tighter relationships between our industry and the academic institutions that educate our future engineers and innovators in an even more relevant manner. If universities, even ones with few resources, worked more closely with industry, companies could simultaneously provide rich problems that have a closer-term impact to industry while academia obtains industry’s knowledge, inputs and tools at an acceptable cost.
Rather than seeing industry as a 100% separate entity from academia, universities should see themselves as a means to educate students on industry’s problems today while performing research that will solve industry’s future problems. The key enabler is to have processes and people in place to link their educators with industry practitioners. It’s not about how much money a university has but rather the personal relationships these schools foster between their professors and industry practitioners to solve complex problems.
I think all universities doing research in our space should consider this if they want to create more research that is useful to today’s practitioners.