Design And Verification For An Era Of A Trillion Devices

Excitement about the possibilities of IoT is balanced by concerns about security, safety, and ethics.

popularity

Scared or excited? When I did a back-of-the-envelope calculation whether the one trillion devices that Softbank’s CEO Masa-san predicted least year at ARM TechCon was possible, I realized that a trillion may be the low end of the range. For me, the geeky excitement about the potential technological progress and how to architect the Internet of Things (IoT) gets balanced very fast with concerns about security, safety and ethics. We as engineers have huge responsibility to get it right in the next 20 years, otherwise we may have an excitingly scary future ahead.

There is this quote by Owen Wilson’s character Oscar in the Bruce Willis movie “Armageddon” as they are strapped into the spaceship to save the world from terminal extinction: “I’m great, I got that ‘excited/scared’ feeling. Like 98% excited, 2% scared. Or maybe it’s more. It could be, it could be 98% scared, 2% excited but that’s what makes it so intense, it’s so – confused. I can’t really figure it out.”

Personally, I feel like that about at least two things these days – being a parent of my 12-year-old daughter who is becoming a teenager way faster than I would have ever expected, and aspects of safety, security and ethics of the Internet of Things (IoT) as we are approaching a trillion devices over the next 20 years.

A trillion devices? Sounds like a lot? It turns out it isn’t, depending of course what one defines as a thing. By 2035, the world population is expected to approach 9 billion, so we need about 113 devices per person to reach a trillion. For me, I am getting easily to a 100 devices either connected today or to be connected eventually. It starts with my 30 or so light switches and bulbs, an alarm system with three cameras and an additional 7 sensors throughout various rooms, 10 connected sprinkler heads making sure I don’t over-water and my solar system with its associated sensors. Then there are daily use items like my fitness tracker, step tracker, sleep tracker, phone, tablet and headsets, tens of thing-like devices in my car and even on my bicycle plus additional devices for PG&E wireless data collection, my “Ring” doorbell and my home wireless network featuring one network switch to get video around and three routers for different areas of the house. My fridge is looking forward to being connected so that it can give me updates on bread and milk. My stove would like to await remote instructions when to turn on, too. Now let’s add in all things that can regulate traffic, can be used in industrial production, can become health implants, and a trillion in 20 years is definitely in imaginable reach.

The technological progress makes me happy to be alive right now and I am an immediate state of 98% excited, 2% scared. There are so many interesting challenges to solve.

Where to process all the data, for example, will be an exciting IoT architecture question to work on. Transmitting all the data into the cloud and doing the processing there is not an option. This Google example shows the dilemma – three minutes of voice recognition by all Android users would require Google to double data centers. So processing has to happen thoughout the path as you follow the data. I went back and looked at Chris Rowen’s brilliant keynote at CDNLive, “Follow the Data.” Process data as you go.

Energy consumption will determine a lot of the decisions, as transmitting the data has energy limits. This will drive new inventions in the area of low power design. For more information, check out the great tutorial on low power design approaches for IoT by Prof. Massimo Alioto at HotChips 26. And in EDA we are helping with design productivity.

Also, those trillion devices want to be developed and if we do not increase productivity dramatically, we will for sure run out of engineers. Technologies like we showed with ARM at DAC 2017 – connecting ARM Socrates  with the Cadence Verification Suite for verification (see Jim Wallace’s blog on “The Future of Tooling“) are only the beginning. Chip design, especially for less complex designs, will have to become a lot easier over the next 20 years.

So geek out we can! It’s an engineer’s heaven and our job is safe for at least 20 years to come.

What could go wrong?

Well, safety and security give me nightmares. A trillion devices are literally a hacker’s paradise. When I asked the gentleman who installed my solar system and said “plug this cable here into your wired home network” about what security mechanism would be in place so that this is not a potential intrusion, he stared at me with blank eyes. Now extend that to the total shut down of infrastructure as shown in another Bruce Willis movie, “Live Free or Die Hard.” OK, 98% scared, 2% excited it is. Luckily, the industry is working actively on flows for security and functional safety from a tools perspective. Interoperability of security architectures in IoT will be key going forward.

This Monday, the roller coaster went for me back to 98% excited, 2% scared when I watched Dr. Phillip Alvelda, now CEO of Cortical.ai and former DARPA program manager, give his keynote at HotChips. Titled “The Direct Human/Machine Interface and Hints of a General Artificial Intelligence,” Alveda gave real life examples of how we have progressed over just the last 10 years in interfacing to the brain. We are now reading out the synapses of mice to understand where they think they are in a maze. The serial cable that has been used for FDA-approved implants is being replaced by optical connections with much higher data bandwidth. We can study and read out the brain and nerve activity of monkeys as they walk in a treadmill – as shown in the picture below.

You will better understand my current 98% excited and 2% scared feeling when I tell you that my dad has a cochlear implant – a very advanced hearing aid with parts of it implanted directly within the skull connecting to his hearing nerves. It has given us back the opportunity to efficiently communicate again. Very exciting! The 2% scared comes in, as my brother and I frequently joke with him, that he is officially the first cyborg in the family.

But where does all this end? The sensors become actuators as well and we can change the state of a nerve, re-creating sensory functions. Combined with machine learning, machines may soon be able to predict how we may react to what they do and say. Consciousness? And talking of AI, as my last movie reference for this blog, the dilemma in “I, Robot” is becoming an actual problem today. Will Smith’s character is upset because during an accident a robot saved him over a young girl because his chances of survival in this situation happened to be better. Still, was that the right decision from an ethics perspective? As we are teaching cars with AI and machine learning how to drive, if a collision becomes unavoidable, are they better off to harm the driver, aim for a pet or a pedestrian? These are ethics questions to be sorted out, and it won’t be easy to do.

So back to 98% scared, 2% excited I am. Or is it the other way around because of all the good things we can do? It’s a roller coaster. As an engineering community, we will have to make responsible decisions in development, and if we do we are enabling a bright future. We are lucky to be alive right now!

What do you think?