The Real Dangers of Assisted and Augmented Reality


smart-car.jpgAnyone who knows me knows that I am a big fan of using science, technology, and analysis to drive continuous improvement and efficiency. I can't help but calculate the time, cost, and relative value of different approaches. For example, "What if we switched to using just one kind of LED light bulb throughout the house and ordered dozens at once when they are on sale? Does that make it easier to replace them? Is the NPV of investing the capital in surplus discounted light bulb stocks now greater than the future cost and effort of making subsequent acquisitions?" This is a real thought process that I have had, debated, and defended recently. Aren't I fun to live with?

I also value self-reliance and simplicity and the ability to take apart mechanical things and repair them. In the last week I've disassembled and fixed our leaking Subzero refrigerator (blocked drain on the condenser drip pan) and our also leaking Frigidaire clothes washing machine (loose connection on the internal filler hose). I love my 40+ year old pickup truck and my 50+ year old Mustang all the more for their repairability. In general, parts are accessible, understandable, and often easily cajoled back into operation (like the corroded connection that occasionally reduces the alternator's ability to charge things). Bonus: Should I need them, replacement parts are often 10x cheaper than their modern equivalents, like $27 for a refurbished alternator for my old Mustang, versus $270 for the one on a modern BMW.  Kids love to ride in the old vehicles, too. "Wow, look I can turn this crank to raise and lower the windows, so cool!" To me, the understandability is a key part of the pleasure. With little effort you can see what each thing is doing and infer how it is doing it. My friends and I may not be able to design a carburetor, but we can take one apart and rebuild it.

So, I'm struggling a bit with the idea of applying machine learning and artificial intelligence technologies to everything around me. Google recently gave me a Google Home device; I'm not quite sure why. Maybe they wanted to give it more real world voice recognition training opportunities. Maybe they wanted me to write about it, as good social marketing. Maybe they wanted to hear how I'm advising their competitors, naughty, naughty. I don't like the idea of it listening to my conversations. My wife really doesn't like it listening to her. The kids adore it. It's something that I feel a natural affinity for anyway, having spent some time at a "smart home" startup a decade ago, which we'd now label "IoT" technology. Controls for our system were amazing stuff at the time, "I can text my home to check on the alarm system or control lights!" They were also frighteningly rudimentary: I had to TEXT my home to turn on my alarm system or control lights because "smartphones" and their app developer tools didn't exist yet. Now I can speak aloud and call up any song I want and have it play in any room, and read the actual lyrics, no matter how garbled the singer. Mondegreens are a relic of the past. I can ask "Hey, Google, why isn't my car charging?" and it suggests possible problems with the alternator. Then it gives me a link to a website for more information or YouTube videos showing how to do it. This is amazing stuff, available commercially today. So much more rewarding than scheduling an appointment with an "expert" mechanic at a dealership who may or may not get to the same fix, but will certainly charge me $150 per hour while he Googles it himself and then marks up parts prices by 50% or more. The point here (yes, there was a point here) is that technology has made it far, far easier and cheaper for anyone to access knowledge and learn how to do things. We can readily disintermediate supply chains and services. Anyone can now be an expert on any topic, on demand. This assisted or augmented reality for everyone is practically magic.
And that's part of the problem, too. Magic isn't transparent. It can help me, it can do stuff for me, it can amaze me, but I can't readily understand how. With new cars, a lot of the systems are too complex and tools too specialized for me to use them anyway. Could I debug the controls on a smart fridge or washing machine? How does Google's natural language processing and search hear my question and find the right answers? What if I disagree or that answer isn't quite right for my particular situation? Do I trust a machine learning model I can't see? Can I ever work out how a deep learning system came to predict a probable cause? Have I lost something by not working it out for myself? Will I eventually lose the ability to work it out for myself? Does it give a vendor too much control over what answers I receive? Will it begin to intrude and guide me to someone's (or some machine's) preferred choice, whether or not that's the best option for me? As artificial intelligence rapidly becomes real world magic, does this help humanity be better? Certainly we can be more efficient. Often we gain insights we couldn't readily work out on our own. But do we perhaps lose something too?
I am frequently irritated by Microsoft Office's grammar checker. A more varied or verbose writing style is not an error; it can communicate more thoughtful nuance or more nuanced thought. There is room for style differences without being grammatically incorrect. Don't even get me started on the iPhone's auto-miscorrect for spelling, or Siri's too creative interpretations of what I just said. Sometimes I feel these "helper" tools are taking away our ability to think and write and use words just how we like. I further dislike how they turn on these functions by default with product updates, forcing me to find the setting to turn it back off again.
To be fair, I also dislike cruise control, anti-lock brakes, traction control, and similar nanny features in cars. Self-driving is an immanently inspiring capability and makes me mourn for a soon-to-be lost era of autonomy and privacy. An inscrutable set of algorithms directs me or even literally drives me to McDonald's. Is it because McDonald's paid more for the ad and a predictive algorithm says most people prefer predictable brands and efficiency-driven prices over quirkier local flavor? Let me write my sentences, let me drive my car, let me discover where to eat. Is it just laziness to accede unthinkingly to someone else's proprietary, encoded, biased opinion? Does assisted or augmented reality really help? Can I get to new analytics insights? Or are we doubling down on dumbing down?
As a wise friend once told me, "It only costs a little more to do it yourself." I love democratized analytics and improved efficiency, but don't augment my reality too far, please. Leave me a choice.
Topics: Internet of Things Data Platforms, Analytics, & AI