USI 2012 Second And Last Day Wrap-Up

The "Universite du SI" 2012 conference is over, and it was fantastic. Awesome speakers, inspiring talks, new horizons, great organization... Here is an extract of my notes about the talks I attended today (after my previous post about the first day talks).USI 2012 Second And Last Day Wrap-Up

"The rise of cognitive Computing", by Stephen Gold (IBM)

The first talk in the morning exposed the motivations and technology behind Watson, IBM's new supercomputer. For those who never heard about Watson, it managed to beat human competitors during a Jeopardy game back in February 2011. The challenge for Watson is to process massive amounts of unstructured data in natural language to perform a probabilistic and motivated guess on a particular question in less than a few seconds. Watson is a learning machine based on Big Data. Stephen Gold didn't say much about its hardware and software. Possible usage start in healthcare or finance. Proposed as a "cognitive-computer-as-a-service" (my term) and run from the cloud, Watson still has no business model, speaks only English, and left open some very crucial questions (HAL 9000, anybody?).

The talk included a demo, which looked like science fiction - except it was real. A physician asked Watson to help diagnose a patient with cancer, and to suggest the best medical protocol. Watson knew how to ask more questions, process results from medical tests, integrate new evidence, and correct its own diagnostic - all that in natural language. It this is the future of computing, then all the programmers in the world need to look for another job.

Interesting facts:

  • 90% of the world's data was created in the last 2 years
  • 12 TeraBytes of tweets are written every day

Favorite quote: "Watson never forgets".

"More than Moore", by Loïc Liétar (STMicroelectronics)

To compete with Moore's law, semiconductor manufacturers need larger and larger investments, and this only leads to pure computing power. Except that customers don't want more power, but new, specialized chips that provide new capabilities (like accelerometers, capacitive touch screens, ultra low power chipsets, etc). ST MicroElectronics cannot compete with Intel on the x86 architecture, but they are leaders on microelectromechanical systems (MEMS). So they don't try to compete with the Moore law, they do "More than Moore" (I'd rather say they work "aside of Moore").

The new components they produce include sensors, actuators (able to provide feedback to the environment), processors, modems, batteries, and even energy harvesters. They must be miniaturized, autonomous, secure, and findable (you don't want to loose your 2mm compass, so it would better be included in a larger object). Large computing power isn't necessary, there is enough silicon already (in your iPhone, in the cloud) to process the data from these components. The main challenge is that, as these new components multiply, the amount of data to process increases dramatically. Big Data, here you come again.

Overall interesting talk, where I realized that consumer electronics are made possible by specialization and miniaturization of existing pieces of technology, rather than pure innovation.

Interesting facts:

  • The semiconductor market was multiplied by 50 in the last 25 years
  • It takes 15 years to introduce a new material in the semiconductor industry
  • Chip makers know how to miniaturize down to 8nm (the current generation of chipset uses 22nm)

"What can we do about parallel programming?", by Bruno Boucard

Parallel Programming is not that hard once you know the strategies, design patterns, algorithms, and implementations. At least, that's what Bruno tried to demonstrate in his talk, with actual pieces of C++ and C# code. A very practical conference, using the classic matrix multiplication problem as an example of how to divide the computing time by a factor of 1000, by leveraging parallel computing (and the GPU).

The domain seems possible to understand, however one hour is just enough to scratch the surface. Also, parallel programming mostly concerns compiled languages. Languages using a JVM (or Node.js) can't take full advantage of multithreading and multicores and fast cache memory. Also, asynchronous and parallel aren't the same thing (but you already know that). The talk was a good introduction to parallel programming design patterns (which have been there for a long time), and in my opinion Bruno Boucard demystified the complexity of the subject.

Definitely something to look into in the future.

"The Elements of User Experience", by Jesse James Garret (Adaptative Path)

Jesse James Garret is already famous (he coined the term "AJAX" in 2005), and his ideas on user experience are now almost mainstream. Putting user experience at the center of product development isn't as revolutionary today as it were in 2001, and you've probably already seen this web design cheat sheet before. JJG didn't spend much time on how to build good products (he's written books about it), but rather he tried to explain what good products are made of (Steve Jobs quotes every two slides). Interesting talk nonetheless, that I took as a refresher course.

Favorite quote: "Products are people too".

"Sense of work, happiness and motivation", by André Comte-Sponville (Philosopher)

My highlight of the day, André Comte-Sponville's lesson was an eye-opener. I love to listen to philosophers who are also great speakers, and who know how to entertain the audience. This lesson (rather than a conference - no slides involved here) called Platon, Schopenhauer and Spinoza to the rescue to show the path of good management (and marketing). Needless to say, it's not enough to fulfill the workers needs to ensure that they do a great job. You have to make sure they love what they do for that (that's called "motivation" in business books, but "desire" in Philosophy).

It would be an insult to try to summarize his talk, so I can only recommend to watch the USI website for the video of his conference. If you have one hour in front of you, I can't see any better way to spend it.

Favorite quote: "A manager is a professional of his fellow workers' desires"

"Blackberry Hill, Domino effect", by Yves Morieux (BCG)

This is the conference that made me appreciate the BCG again (not like yesterday's talk). Yves Morieux has rage. He sees the flaws in our organizations, he sees why our businesses cannot go on with productivity gains. He understands why the classic "strategic alignment" process supported by Michael Porter theories doesn't apply to today's companies. During his talk, Yves explained that this process leads us to respond to complexity by complication. But since our competitive environment becomes more and more complex, it makes our organizations more and more complicated. That's why we spend most of our work time not doing actual work, but rather fighting with (or against) the organization. That explains the disengagement of workforces. That also casts a deadly shadow on our future, because the complexity will go on increasing. Except if we change our strategies to emphasize cooperation rather than processes, matrix management, committees, incentives and the likes. And the CIOs have a central role to play in this shift.

Excellent talk, highlighting the infuriating mistakes that top executives keep on making. Makes every CIO/CTO in large companies want to shout during the board of directors. For smaller companies (like mine), the same principles apply, but to a lesser degree. All in all, Yves Morieux's talk brought me motivation and a clearer view of the (nefarious) consequences of the strategic alignment principles.

Interesting facts:

  • 3% growth per year = we are twice as rich as our parents
  • Only 11% to 23% of workers feel engaged at work
  • Complexity was multiplied by 6 in 55 years

Favorite quote: "Improve knowledge of others".

"When Computers become empathic", by Rosalind Picard (MIT)

Computer can now, to a certain degree, read emotions. They can't do much with this information, except maybe explain these emotions to human beings who lost the ability to decode them (like in autism). The MIT research on that field seems to be at its infancy. If we can now log the stress level of any individual using a watch-like bracelet and a bluetooth connection, it looks like we're still far from interacting with empathic computers.

"Different ways for design", by Philippe Starck

The famous designer made the closing keynote in a personal, exuberant, chambolic kind of way. He piled up a lot of ideas that he explored during his career on design. From "convergence" to "democratic design", illustrated by pictures of objects he designed, the talk was pleasant but not exceptional. It's good to see that Starck put some very precise motivations before every design, and that these motivations tend towards a better world. It's also good to hear about degrowth and ccology in an IT conference. And it's a good motivation to be more creative and follow your own convictions in your work.

Favorite quote: "There is never any product without a vision".


Octo Technology scored very high in my "top conference organizers" list. I learned a lot, discovered new theories and patterns, met interesting people, and most of all I came back home with a head full of ideas for my day-to-day work. Not to mention the final concert by Irma, which was a great pleasure to listen to.

I'll come back next year for sure!

Published on 26 Jun 2012 with tags conference management not technical

comments powered by Disqus