I had the opportunity to hear Dr. John R. Williams give a talk titled, "Talking to Light Bulbs-- How the Internet of Things is transforming everything." This was a particular treat, listening to the man at the bleeding edge of innovation in my field give context and perspective to the IoT movement revolutionizing modern life. This is my attempt at setting down the points that resonated during that speech.
Right now, the best taxi service in the world doesn't own cars; the best retail service in the world doesn't have a physical location. And the best hotel service, doesn't own a single hotel. Data is being generated at an exponential pace, and analyzing this data is enabling the above to happen - it's even helping teach farmers how to be better at their job, beyond their own personal experience.
The beginnings for this were set up with the founding of the World Wide Web. Originally just to be used by the United States military to keep track of their stuff, the purpose soon expanded. Another milestone was in the development of HTTP - according to Dr. Williams at the time because HTTP was text-based it could get past firewalls since text was deemed safe; this made HTTP more useful than binary protocols.
The Internet was intended to support documents only. The coming about of Javascript birthed the creature known as a web application, which soon supplanted that original vision; it enabled developers to build programs on browser engines, which meant that these web applications can run on whatever device can run a browser. Now web applications are ubiquitous - practically all web pages aren't static documents, but rather are code that aggregate information from a multitude of sources. The code also has a second ability - it can track the user, and send out this information.
The revolution has only been helped by the open-source movement and the practice of releasing API's; developers can mash up different technologies and come up with novel applications. One example of this that Dr. Williams showcased was an app created by his students - the app leveraged the Meetup API to provide real-time information about meetings, filtered to those meetings comprising more than fifty people. The point of the app was to provide taxis with information on where to go for possible fares.
The coolest thing though that was demoed during the talk were the talking light bulbs. This was connected lighting - light bulbs able to change hue to create an atmosphere, able to turn itself on to wake you up, able to be toggled into security mode from your phone. Lights that can turn themselves off and on again to notify you of an emergency or a phone call. Light bulbs that turn on as you approach your house.
The Internet of Things isn't all upside. It has proved to be a very disruptive force - many old businesses do not exist now because they have been made obsolete. Some companies have found success by embracing these changes and adapting to the changing times. The problem is that companies grow so successful they think they are too big to fail - it takes a special effort to always be on the lookout for the next sea change. Companies need to always be observing and orienting, and they need to decide and act quickly. In software development, this methodology is embodied in Agile programming - which in turn owes a lot to John Boyd and his OODA loop.
But IoT isn't just dangerous for companies that don't adapt; in many ways it's dangerous for its proponents as well. With increased connectivity and generated data and the current trend toward analytics, many people fall into the trap of thinking that the sky's the limit, that they can predict anything, and that the more information they put in the more in control they are.
This is understandable because for the longest time the prevailing mindset was that we can create mental models of phenomena to understand them. As more information comes up that shows minor problems with our mental models, we can make changes accordingly.
But what if a single infinitesimal change resulted in behavior completely different from that predicted by the model? The discovery of chaos and the butterfly effect mean that very small initial differences can result in huge perturbations. Dr. Williams illustrated this with the three-body problem - while it may be easy to predict the movement of a body hanging off another like a pendulum, adding a third body to the pendulum bob just makes predicting movement a hopeless endeavor.
Dr. Williams also related an anecdote about his colleague who was trying to predict the weather. During his calculations, the colleague made a mistake on the digit at the fifth decimal place; the result was that all his model's predictions were wrong. The colleague concluded it was impossible to predict the weather, because our measuring tools will never be accurate enough. And that's without taking into account the tools changing what they're observing as the measurements are being done.
Even in analyzing data one doesn't make use of all of the data; Dr. Williams mentioned that in order to make predictions often they just look at a fraction of the data - because data is almost always dirty, and making the algorithm fit the data too closely makes it too sensitive to these minute changes we've been discussing that result in wildly off predictions.
So there have to be boundaries. Within those boundaries, we have to allow for patterns to emerge on their own, perhaps introduce new strange attractors or dampen non-helpful ones. Of particular help would be the Cynefin framework developed by Dave Snowden.
Overall, it was a great talk. The above video was also used by Dr. Williams, but I couldn't find the other material I remember he used. I find some similarity with N. Taleb's ideas; lots of stuff to ponder.
