Sunday, February 27, 2011

What Are You Waiting For ?

Bob Cringely reminded me of this defining moment when two DEC executives looked at a dismantled PC on their desk:
...But the most telling story about Olsen that Avram tells isn’t in his blog.  It was about how he and Ken Olsen bought one of the first IBM PCs and disassembled it on a table in Olsen’s office.
“He was amazed at the crappy power supply,” Avram said, “that it was so puny.  Olsen thought that if IBM used such poor engineering then Digital didn’t have anything to worry about.”
Clearly Olsen was wrong.
I find this little story very telling because it shows Olsen in 1980 very much stuck in the late 1960s when it came to what mattered and what didn’t in a computer.  Yes, having a robust power supply was good for computer reliability, but not as important as having a great operating system and applications. But that’s not the way Olsen saw it.  In a world that had to that point been dominated by computer companies building expensive products aimed primarily at engineers, he simply had no concept of a computer as a consumer product....

What would somebody making expensive hyperspectral imagers say about a webcam and some prisms ? what would an optical engineer say about a microscope with no lens ? Very likely something along the lines of what was said in that DEC office. I have actually asked that question to some specialists, a shrug is a nice way of framing their answers. Not to avail I like being the candid ones in these discussions. But going back to the issue of making it in the real world, one has to face the unbearable sluggishness of how things move along.  More precisely, Jerry Weinberg pointed out that

Most of the time, for most of the world, no matter how hard people work at it, nothing of significance happens.

And as much as what Hollywood wants to convince you, in the journey of maturing a technology there are few aha moments. You're in the dip and the reality of it is that most technologies take some extreme combativeness to be born in reality. More precisely, most of them don't survive the climb on the TRL scale.

When John Mankins created a scale of Technology Readiness Levels, he was mainly trying to rationalize the different investments NASA was undertaking in a wide series of technologies. Now we know that NASA funds more in the realm of TRL 3-6 for its R\&;D, while other agencies like the NSF focus on TRL 0-1 and maybe 2 (if you're lucky). John did not mention TRL 0 as it could be a placeholder for theoretical work that does not precisely hinges on an actual technology. Using this scale here is how I see the life and death of technologies and some examples related to Compressive Sensing.

All the theoretical work fits into the left hand side of the graph. In order to prove to the world that CS is real, you can always point to both the TRL 9 MRI example or the Herschel PACS camera. However, those examples are just accidents in that either the space camera was built to be underused in the first place or that in MRI you were already sampling in the right phase space. Aside from these accidents, one has to push things to the right. Most other technologies described in the hardware page go from 1 to 3-4 and some won't survive the process because they are in direct competition with other technologies. Let's take the example of the single pixel camera. One of the reasoning for using it is that the detecting piece is expensive (say IR or Teraherz sensors), some technology breakthrough could change that and make irrelevant a different scanning systems such as that enabled by compressive sensing. What is truly amazing with compressive sensing is the ability to use an extraordinary set of tools such as the many reconstruction solvers, dictionary learning and the Donoho-Tanner phase transition to see if there is a fighting chance that the technology can go the next level.  What are you waiting for ?

No comments: