Autonomous cars not around the corner

April 26, 2017

KITT_Knight_Rider2The car industry has its challenges. From being the ultimate provider of freedom, personal transportation and quality of life it now is supposed to solve the safety and sustainability issues it created. The very complicated and expensive cars of today are not utilised enough (5,5% according to my TeliaSense app), consume too much space, cause a lot of accidents and impact our planet badly when produced, used and discarded. It’s easy to suggest public transportation as an alternative but as soon as leaving the urban areas that is falling short. Intuitively the solution would include new ways of owning and using cars, electrical engines, even more safety features and cross-industry innovation within the car eco system. And since we can’t wait 10-15 years the cars already on the road must be included at least to some degree.

With this in mind I find today’s focus on self driving cars strange. We have continuous safety improvements in new cars, they are connected and loaded with sensors and most manufacturers have electric engines in some models. Also cars on the roads are being connected using the OBD port and offered services from the car eco-systems. But autonomous cars is what the industry, media and strangely enough IT companies are talking most about. I understand that manufacturers work on self driving cars since it’s a very complex challenge which will take many years to sort out, and it will bring continuous innovation to the cars down the road. But why talking so much about it already today? The technical challenges are big but not what will determine how soon we will see self-driving cars on the road. It’s culture, law and policies!

Take a look at airplanes. They are very complicated to make and fly but self-flying planes are here or at least around the corner depending on definitions. So are we ready for choosing flights with or without pilots? And which one would be cheaper? Imagine the first autonomous car running over a person or two in US, and what the penalties will do to the car maker involved. Some people argue we will have new infrastructure for self driving cars. But with new infrastructure I guess we could be more innovative than making cars without steering wheel.

The SAE (Society of Automotive Engineers) classification for autonomous cars has six levels, from none (Level 0) to fully automated systems (Level 5). Level 3 is a vehicle in which within known, limited environments (such as freeways), the driver can safely turn their attention away from driving tasks, but must still be prepared to take control when needed. Berg Insight estimates registrations of 16,5M new Level 3 cars and 7,8M new Level 4 cars and no fully autonomous ones (Level 5) 2030.

I’m sure we can make very good but not perfect self-driving vehicles a couple of years from now but can’t see them being used on public roads for many years, unless very limited in speed. There has been some early progress when it comes to regulation here and there but I haven’t seen any progress in the field of liability yet. And I guess a generation or two of drivers will probably have to disappear from the roads before it happens.


Intelligence in the edge

September 10, 2015

ai-cropped-640x353After the Billions-of-devices-hype came even bigger numbers on the Big Data opportunity. It’s obviously so that being able to gather and analyse huge amount of data from different sources gives interesting opportunities for both good and bad guys. And the big number game always works: if airlines save 1% cost it’s a huge amount of money. Or if rescue shows up at car accidents a minute earlier on average we save many lives and loads of pain and money. If people on the planet save a minute a day on their way to work we would create mountains of time to be used for something fantastic. We have left the Billions and talk Trillions now.

Let’s come back to the view that enormous shared data power in the cloud is the only way forward. We’ve heard it before. It all started with IBM’s mighty mainframes in the core and 3270 terminals in the edge. Computing was only about central expensive machines until Apple turned it all upside down with the first personal computer. And ever since have IT managers and others had to struggle with the balance between edge and core. At 3Com we invented Ethernet and argued power to the edge. Why send data outside the office if it’s internal and only to be shared between people in the office? It’s at least not safer and what if connection to the center is lost? It has been going on like this, back and forth, Thin Clients with all power in the core was really the future for a while and most recently employees started bring their own, rarely thin, devices to work.

There is definitely a lot of economies of scale to run things from the center. But innovation often comes from the edge. That’s why we have regional and local governments for example. Culture, climate, economy, religion, desire and everything else differs from place to place, company to company, human to human.

Coming back to Big Data and IoT. First of all, IoT is most often small data, and little or no velocity, variation and all the other V’s that Big Data is supposed to deal with. Secondly, in many applications we need to make really quick decisions down there at the floor, in the edge. A home care solution which tries to identify when a person living there is about to get stroke, requires a thorough understanding of that individual, constant monitoring and learning and very rapid and correct alarms when something is about to go wrong. A retailer who need to put the new pasta somewhere need to make that decision now. With real-time support for that decision it will be a better one than without.

I believe a new wave of distributed AI applications within self-learning systems will be a really important part of the Internet of Everything. Keep an eye at companies like our THINGS member Aifloo and Imagimob, both developing AI capable edge devices for industrial, healthcare, retail and energy markets. These clever devices with their own sensors built-in, might even make it possible to avoid further integration into larger equipment to provide applications requested.


%d bloggers like this: