The first keynote on "killer apps" addresses safety failures of the US Army's Patriot missile, the Therac-25 anti-cancer radiation system, and a leading automaker's vehicle systems. Testing for errors in accuracy, calibration, and control must cover continuous use rather than brief iterations of activity. Turning something on briefly does not constitute a use case. Design engineers too often assume that thousands of hours of iterative laboratory testing are substitutes for real-world use that continually stresses tolerances. I was not surprised to learn that some government regulators lack the in-house software expertise to evaluate control system failures.
I talked my way into the International Engineering Consortium (IEC) Management Summit because frankly I'm the kind of guy who belongs at summits. Much of what the high-level participants discussed was off the record but my own opinions about system engineering are very much on the record. Get ready; here it comes. My vision of the Internet of things (IoT) is a biosphere of multiple ecosystems optimizing human processes that have never before been analyzed or controlled. Engineers shouldn't dream up IoT devices without input from their marketing team to determine whether some IoT-enabled thing will sell. There is a huge IoT opportunity in legacy pre-Internet physical infrastructure whose industrial control systems were never designed to use networks and embedded sensors. Young software engineers (20-30yo punks and whipper snappers) must converse with the retired mechanical and electrical engineers (60-70yo fogies and geezers) who designed those legacy systems. Intergenerational knowledge transfer will enable IoT systems to address those legacy facilities' gaps in performance and security.
I really liked Max "The Magnificent" Maxfield's morning session on the possibility of a robot uprising. This is what serious engineers spend their time preparing to handle. I said yes to the free breakfast (the program guide said so) but I didn't put on a tinfoil hat. Plenty of conference alumni wore their tinfoil hats, as you can see below.
I have no idea what the screenshot of giant singing jellybeans had to do with the topics of AI self-awareness and pervasive automation. The engineers present had no consensus on design protocols that could mitigate a spontaneous robot uprising. Isaac Asimov's Three Laws of Robotics should be the obvious solution given the attendees' penchant for revisiting their favorite science fiction stories.
The second keynote on open source hardware and embedded systems was from an MIT PhD nicknamed "bunnie." I didn't know genius hackers could have innocent nicknames but when you're Dr. Andrew "bunnie" Huang you can do anything you like. He discusses reverse engineering of common hardware on his blog. His insight into the slowing of Moore's Law opens longer product development windows for hardware. The last node of Moore's Law will not allow for further reductions in computing cost. We should welcome the end of planned obsolescence if it leads to bunnie's vision of repair and recycle cultures for legacy hardware. Open-source hardware enthusiasts like bunnie have as much fun cracking their own homemade PCs as open-source software developers have with free products like LibreOffice. Gearheads in tech can crowdfund their latest hardware hacks at Crowd Supply because the world needs more DIY solutions. It makes sense for bunnie to be a leader in the maker movement. I think makers should prepare for the coming of quantum computing, which will blow away the Moore's Law limiting nodes.
I spent much of day two at the Silicon Valley Open Innovation Summit. A lot of the presentations were sales pitches from larger firms and debut pitches from startups. It was enough for me to see new tech like flexible batteries. The lead sponsor proposed that common building blocks for innovation in a few key areas will cross all product lines. Okay, but please find academic support for that contention. MIT's Sloan Management Review found that innovation has more to do with human cultural attitudes than specific technologies.
I learned a bunch of new terms at this innovation summit. Metadesign should make collaboration easier. Someone mentioned "technology arbitrage" but I am unable to locate a definition; I take it to be the temporary advantage a niche maker possesses before competitors flood into a market with cheaper processes. IoT device makers and service providers are segmenting their markets into "brownfield" and "greenfield" parts, and the legacy infrastructure I mentioned above falls into the brownfield category. The sectors that have little coverage in research literature, like flexible and printable batteries, appear to offer first-mover advantages for startups.
The best show on the expo floor was the Fantastical Theatre of Engineering Innovation. The prevalence of cotton candy and popcorn enhanced the circus-sideshow approach of several exhibitors. Some of them were more performers than exhibitors, with crawling robots and flashing thingies scampering all over the place.
Max Maxfield's "beer and bacon" talk stole the show. I missed the bacon and I didn't need the beer. He's making progress on a prognostication engine that predicts women's emotions based on inputs from environmental influences and a man's decision points. Giver her flowers and a dial moves. Some dial lights up during a blue moon. Wow. This is how hard-core engineers spend their spare time. Max loves tinkering with Arduino, antique parts, and lighting fixtures. He'd be a cool guy to have as a hardware startup's technical advisor if entrepreneurs can get his attention.
I took away several conclusions from this show that are relevant to the financial sector. Concurrent hardware and software development should be an obvious practice, especially with Moore's Law slowing down. Synch those waterfall charts and show the C-suite how the capex budget accommodates both development cycles. Rapid prototyping requires embedded design, so firms that don't have a metadesign culture need to get cracking on building one. Failing to catch development errors quickly hurts product development. The new product development window (at least for the semiconductor sector) is three to four months long while Moore's Law still holds. Missing this window means losing the chance to deploy an entire product family. Finally, knowing your sector's product life cycle means knowing how to disrupt it, sort of like knowing how to disrupt a competitor's OODA loop. This stuff is pure genius. I'll share it with startups that deserve my early stage investments.