The IT world just moves too fast for Universities.
I have been known to claim that aside from the Cryptography and Artificial Intelligence fields, the net effect on the state of Computer Science as contributed by the academic world is effectively zero. For example, there are hundreds of User Interface Design faculties all around the world (such as My Alma Mater!) and yet the frontrunner in modern Computer-related design doesn’t get any of their UI knowhow from any sort of academic influence. They do not participate in research programs. That’s very odd.
An example has come along to test this observation. If you so happen to be a technogeek or a programmer, you must have been hiding under a very large rock if you haven’t seen this amazing “Content aware image resizing” demo: (If you’re impatient, jump to the middle and end of the video. The shiny stuff really starts when there’s 3:25 left on the clock).
Clearly this reeks of academic involvement. It’s actually mostly an in-house development from MERL, The United States R&D branch of Mitsubishi. However, for arguments sake, let’s say this WAS in fact developed entirely in the research department of some university or other.
Normally this would be PhD material, if you’re lucky enough to find an open minded professor. However, no amount of forward-thinking in a professor is going to allow you to present that one 5 minute demo along with a software program as a thesis for a PhD, eventhough clearly they speak entirely for themselves and is vastly more worthy of PhD status than most actual PhD theses. In other words, in order to turn this brilliant piece of software engineering into an actual CS PhD, the unfortunate authors would have to bunker down for 6 months to write an enormous paper. In CS terms, 6 months is a virtual eternity.
Instead, they released a youtube video. Effects:
- over 300,000 views, presumably focused primarily around programmers and tech people. I have yet to see a research paper other than the one about the RSA asymmetric encryption algorithm that has generated this much attention in CS. That’s crypto which I already excluded from this observation.
- One of the two authors got instantly hired by Adobe because they want this stuff in their next photoshop release.
From photoshop this technology will see exposure far exceeding the 300k viewers it’s already received. I wouldn’t be surprised if content-aware image resizing shows up in FireFox next year. Give it another year or so to trickle down to the rest of the population, which gives us the sum total of 2 years for a completely new piece of technology to become as standard as bread and butter in the CS world.
All by writing some software, crafting a shiny demo, and posting it on youtube.
Even if this was a university effort and there had been a large thesis to go with this video, the thesis part would be unneccessary and wouldn’t have added anything. The authors would have been better off going their own way.
I can’t prove it, by my instincts, and vast amounts of recent Computer Science history, tell me that this pattern repeats for virtually every CS improvement (outside of Crypto and AI) ever. So, for those of you still making the mistake of getting a CS degree:
If you have balls, quit the moment you have a good idea. Universities are not capable of nurturing it anyway.
Kars http://leapfrog.nl
August 29th, 2007But Reinier, what about the knowledge building that is facilitated by those theses? I’m only half joking, but don’t you think sometime we need a little more than just a YouTube video and source code to build on someone’s ideas?
Charl Botha http://cpbotha.net/
August 29th, 2007Okay, I had to bite.
You make far-reaching and provocative statements with surprisingly flimsy evidence or argumentation. I’ll mention a few points:
1. Your alma mater is perhaps not the best example of groups currently on the cutting edge of human computer interaction. Also, where do you think many of Apple’s (or Google’s, or Microsoft’s) brightest employees come from? That’s right, judging by all their PhDs, academic institutions!
2. Content-aware image resizing was fist and foremost presented at SIGGRAPH 2007 as an academic research paper, see http://www.siggraph.org/s2007/main.php?f=attendees&p=papers&s=3
This would be perfect as a very significant part of a PhD thesis, had it been developed by a PhD student. In this case, it was developed by a post-doctoral research scientist.
3. Dr. Ariel Shamir, second author, has a double affiliation: one as research scientist with MERL, and one as faculty of an academic institution: http://www.faculty.idc.ac.il/arik/
4. The same is true of Dr. Shai Avidan, first author: http://www.faculty.idc.ac.il/avidan/
5. Both of the authors have academic degrees (PhDs). Go figure. They did not have the balls to quit their studies!
6. MERL, or Mitsubishi Electric Research Laboratories, is a research institute with a clear academic / scientific emphasis. Some of my friends have spent part of their PhD-projects there, I know of a number of academic research projects (Cube4 volume rendering architecture) that have been developed there in cooperation with universities.
7. “All by writing some software, crafting a shiny demo, and posting it on youtube.” - you know that it’s not this simple. These two scientists both have years and years of *academic* experience (and the PhDs and publication trails to show for it). Once again, this was first and foremost a SIGGRAPH publication (very high-impact in the graphics world), the internet popularity came later. The latter is also important, the former is crucial.
I would counter your main idea with the observation that there are a million times more crappy ideas than there are good ideas. So, if you have the balls, take more time to work on yourself and on your idea, until you’re able to elevate it to the level of content-aware resizing. A good research group is not a bad place for nurturing people and blue-sky ideas that might make it big. I do admit that it’s not the only place.
Reinier http://zwitserloot.com
August 29th, 2007Sure, but the ‘more’ needed is a combination of experience and experimentation. I don’t see what Universities have to add to this. Someone with a willingness to learn and an internet connection can get much more done without the vagaries of having to learn a 40 year old broken model (waterfall), a language thoroughly unsuited to learning programming (java).
Perhaps more importantly, the world of CS still consists primarily of low hanging fruits. I would consider this image resize stuff low hanging fruit. You don’t need to understand much about image processing or even Mathematics to follow what’s going on here, and writing your own implementation using just the youtube video can’t be too difficult. Constrast this to cutting edge improvements in the area of Mathematics.
As a repository for gathering and sharing knowledge, Universities are becoming obsolete, as information is at your fingertips these days, regardless of where you are. Universities are moving more towards research and teaching. Teaching is clearly not working, at least not in CS. Even the very basics of teaching rudimentary programming skills is usually bungled. (Java is certainly an acceptable language, but for an introduction? Even Hello World contains at least 10 different difficult concepts which must be waved away and taken on faith for a long time. That’s, didactically speaking, really stupid!)
Research is also not working, by practical observation that for the vast majority of subfields in CS, every single new improvement is developed with no input from any sort of academia.
Charl Botha http://cpbotha.net/
August 29th, 2007Hello?! I posted a fairly long retort this morning, but my comment has not appeared yet. Perhaps eaten by your Akismet? Could you check?
Thanks.
alper http://www.alper.nl
August 29th, 2007Hello Charl,
I fished your high quality comment out of the akismet cesspool. I have no clue what it was doing there. It should appear now.
Reinier http://zwitserloot.com
August 29th, 2007I think you partly nailed me there, Charl, -but-, on the other hand, this image resize stuff got famous off the back of a youtube video. I wonder how fast Ariel Shamir would have gotten a job if this had been a PhD thesis instead.
Apparently, then, universities are still a useful educational tool, but they do not ’set the trend’.
My bias does not stem just from the lack of cutting edge nature of the TU Delft’s MKT department, but also from experiences in academia-based conferences. I was as much part of the problem then, so perhaps I’m being hypocritical, but on the other hand, I learned my presenting skills by being awestruck at the insane quality of presentations at non-academically-organized conferences. And, of course, youtube. They are a dime a dozen on that circuit, whereas in the academic world it’s virtually unheard of.