Resident PHD- holder Regis Behmo explores the idea that, while R&D is constantly paving the way for new inventions, startups, and advancements, it seems that Academia itself could use a little disruption.
You can follow Regis Behmo on twitter at @RegisB
Last November I attended the International Conference on Computer Vision (ICCV), the most prestigious scientific conference in the field, with more than 1500 participants, and 330 accepted papers, which very often redefine the state of the art of a specific application and open up novel research opportunities. The conference lasts for a little more than a week and for computer vision geeks, it’s paradise.
You would think the social networks would be abuzz with comments on every presentation. I expected speakers to respond to questions asked online. There would be controversies and back-and-forth arguments on whether a particular research work was deemed innovative enough to be presented. I thought people would talk about their idea to improve on other’s work. There would be live tweets, long Google+ comments and insightful blog posts.
There were not.
In one week, I counted a grand total of fifty-four tweets related to the conference, fifteen of which by myself. The same observations were made at other famous conferences. Scientists don’t tweet. I read about twenty blog posts a day, and less than one good computer vision-related blog post per week. Scientists don’t blog. I have 327 LinkedIn relations, less than a dozen of them are tenured researchers. Go and try to find David Lowe (inventor of SIFT), Jitendra Malik (Stanford teacher and Computer Vision rock star) or Andrew Ng (Stanford teacher, startup founder and machine learning idol) on your favorite social networks. You will either fail or be disappointed. Scientists don’t network on the web. At least there are Wikipedia pages about them.
Are scientists just uncool geeks?
The social network revolution has surpassed academia without so much as a wave; but this is also true for the other two revolutions that rocked the Web in the past ten years: remember Web 2.0? the wisdom of the crowds and collaborative editing? That didn’t happen, not in Science. What about the move to the Cloud? When I mention Hadoop to a researcher, he says “bless you”. Why is any of this important at all? Because startups have an opportunity to have a significant, positive impact on the way Science, and thus our world, move forward. The age-old peer-reviewing system can be improved by taking into account comments from other, more diversified sources, such as a crowd. Researchers don’t need the opinion of a reviewing committee on whether a paper should be published or not. The $1000 price tag of scientific conferences could be challenged. It is possible to shrink the market while addressing a wider audience.
“Those who have made money in Science, please step forward”
Success stories of startups with customers in the academia are scarce. Mendeley, the article sharing-based community is a notable exception. This lack of interest can probably be tracked back to the fact that business and engineering people know very little about the many worlds of research. Take the above example of cloud computing: research labs do not buy CPU time on Amazon EC2. Instead they buy large computing clusters or GPU farms. These machines will become obsolete after a couple years of use at 10% of their maximum capacity. They are loaded with unpractical software. Grad students don’t have a clue how to administer a cluster. Research teams have to share the cluster and schedule their experiments ahead of time. So why don’t they rent computing instances in the cloud instead? Every researcher I have talked to about this issue has given me the same answer: computing capacity rental is hard to justify on a yearly financial report. Surely, there must be a simple solution to this problem, right?
France’s research institutions are less rigid than we tend to think. The organization of the INRIA(Inventors for the Digital World) revolves around small project teams. The laboratories in the Grandes Ecoles are open to new work methodologies. Grad students have a lot of freedom to decide which tools and methodologies they should be using. Such a market segmentation looks like the perfect starting ground for a lean startup.