IoT sensors breed a new future as technology fades into the background. Old industries are reformed, new applications light up and take flight while data stalls. Tomorrow begins today!
With more than 50 billion devices already in the market, what does the future hold for ARM and the innovation for the next 100 billion ARM-based chips? To find out, Krisztian Flautner, Vice President of R&D at ARM shared his insights with jblyler Chief Content Officer at Extension Media. What follows is a portion of that conversation.
Blyler: Let’s jump right into it. What business and technology trends over the next five years will push ARM well beyond 50 Billion chips?
[Click image to enlarge]
Flautner: Technology will become less and less visible in the future. This is a sign of market success and perhaps maturity, i.e., when you stop thinking about a piece of technology as technology. You take it for granted and instead focus on the experience.
We are at this point today. We have moved beyond the early phases of technology development, where the initial complexity of the devices required a lot of work and attention. The new era will see more and more of these technologies disappear into the background. The Internet of Things (IoT) is the breeding grounds where some of this will happen, delivering the much talked about age of ubiquitous and pervasive computing - we are deploying sensors throughout our environments. When combined with more computing power, these sensors will make the environment adapt to you instead of the other way around. This scenario is starting to happen today and we’ll see continued progress over the next five years.
Blyler: The future with an adaptive, sensor-rich environment where technology disappears into the background. What else can we expect?
Flautner: It’s kind of an oxymoron, but a lot of the exciting things that will be happening in the next five years are very mundane. They are exciting precisely because they are mundane. Computer technology will be used to significantly reform older industries.
Think about the reformation of older technologies with the introduction of computers. There are many traditional industries that have not really leveraged the potential of the modern computer, e.g., high voltage electricity and power distribution, engines, motors, and the like. I think we will see computers changing the landscape of these industries. ARM has several research projects studying ways in which to design motors very differently with deeply embedded microcontrollers. It turns out that fundamental motor design assumptions change dramatically if you can use a powerful dynamic controller.
Blyler: Embedded processors pushing further into the mechanical world to reform an entire industry is quite a story. Is this an active pursuit at ARM?
Flautner: It came about as a genuine research project. We don’t know what the answer is, but it seems like an interesting application. For now, it is a trend that we are watching. There are several power electronic start-ups working to develop sophisticated, microcontroller-based systems to make the production of electricity cheaper. They approach this problem in a fundamentally different way. One example is a Cambridge start-up called Amantys, which is developing new power semiconductor technologies.
Figure: Amantys worked with EMEF to bring state of the art insulated gate bipolar transistor (IGBT) technology to existing locomotives.
Another trend is in lighting. Many ongoing IoT applications are now combining microcontrollers with lights to reduce power consumption and to communicate with user applications.
Robotics is another area where processors play an increasing part. Have you seen that MIT YouTube video (below) about a robot that flew like a bird? A student remotely controlled the flight of a robotic bird that flew by flapping its wings. The movement of this non-linear robot system was incredibly realistic. Controlling this type of dynamic system has only recently been feasible thanks to the availability of embedded computing platforms for highly non-linear applications. In other words, designers now have a very different (non-mechanical) way to control the limbs of the robotic bird. It is no longer just science fiction.
Video: The first autonomous flight of MIT’s Phoenix ornithopter - a substantially redesigned machine with onboard sensing and computation.
Blyler: Non-linear computations require a lot of processing power, but the power consumption for a bird in flight has to be extremely low.
Flautner: Absolutely. The first version of this bird required so much computing power that it had to be off-loaded, requiring a large radio transceiver and extensive remote computing - but, the newer versions use the latest embedded cores and the computations are performed on the bird itself. This opens new opportunities for robotic non-linear control systems.
Blyler: One trend you haven’t mentioned is the one involving big data. The accelerating growth of sensors and the potential reformation of the electrical power industry should result in a tremendous increase in raw data. Many futurists see big data as a key business opportunity. Does ARM see a future in services such as the gathering, analysis and selling of sensor data?
Flautner: I don’t think we are on that path. I recognize why people think there might be a big data type of application around gathering sensor information and analyzing it to predict things based on that data. It could be a valuable business, but not one we are considering.
I also think some of that intuition is wrong. The reason why there is a lot of activity in that area is because most people think big data means '''Google-like" business models, but the existing Google-based landscape is very much based upon relatively mundane data that is used to sell advertisements at a good price. In other words, Google is successful because they have a very easy to understand business model based around understanding something about people and the repackaging of that as advertisement. I’m not actually sure the sensor world will be the same, but that is my opinion.
Blyler: How is the sensor world different?
Flaunter: There will be a lot of value in understanding what the sensors are actually telling you. Some of that may be big data and some of it might be very small amounts of data coming from the right (critical) sensor. Perhaps there will be value at times in some of the sensor data, especially where knowledge about people can be monetized, but I’m not sure how valuable the sensor-based big data market is going to be.
Having said that, I think every time you have access to lots of data the first thing people do is try to figure out what to do with it. It seems inevitable that initially we will see a lot of big data solutions to sensor problems - but over time and with a better understanding of the space, the number of big data solutions may be fewer than expected. It isn't clear to me that big data type applications will be the ultimate business for the sensor space. However, there are a lot of people hanging their business hopes on application for (big) sensor data. We are still in the early stages of this business so it is hard to be certain.
As for ARM, we are much more interested in the little data challenges such as, how do you make sure you can get the data to the cloud? How do you make it so easy to use that people don’t think about it as a problem anymore – e.g., making the technology less visible? How do you build simple sensor devices that easily connect, easily upgrade and can be managed remotely? The answer comes from the ecosystem. An innovative ecosystem will figure out how to best use these sensors and whether to do the back-end processing on a large cloud data store or whether they want to do it in a distributed way in the network itself. I think those will be driven by different criterion and different believe systems.
Blyler: Thank you, Krisztian. You’ve provided the ARM Connected Community with many great ideas as to how the future might be shaped with the next 50 Billion devices. For example, will technology really fade into the background as the IoT movement takes hold? Will that ubiquity be good or bad for users? Which older, traditional industries will be reformed by ever faster and lower power embedded processing technology? What new creations – e.g., the algorithmic intensive robot bird – will become commonplace in the near future?
I look forward to reading the discussions that follow.