The industrial internet

Presentation and video at GE's 'Minds + Machines' event in New York, talking about what GE calls 'the industrial internet'. Pretty much every single piece of machinery in the industrial capital base will have some sort of sensor or network connection. It seems clear that this will change things just as much as much as PCs and enterprise software did in the past.  

Read More

The internet of things

My grandfather could probably have told you how many electric motors he owned. There was one in the car, one in the fridge, one in his drill and so on. 

My father, when I was a child, might have struggled to list all the motors he owned (how many, exactly, are in a car?) but could have told you how many devices were in the house that had a chip in. 

Today, I have no idea how many devices I own with a chip, but I could tell you how many have a network connection. And I doubt my children will know that, in their turn.  

At each of these stages, the things that this new technology would be used for were hard to predict, and many would seem absurd. (A little motor in the wing mirror to adjust it for you? Really?) But the trend is inevitable. Sensors and signals and connections in our lives will proliferate in ways we can’t predict, and in ways that would probably seem pointless if described to us. The cost of a network-connected sensor will be tiny (partly driven by the smartphone supply chain) and the battery will last years (and with solar or energy-harvesting it may never need to be charged), so uses will be found that will seem as natural to our grandchildren as a blender (“Really? You can’t chop by yourself?").

So I’m a big believer in the ‘internet of things’ at a macro level. The question is how we get there and what it might look like. 

One vision is that, ‘of course’ (always a danger sign), all these devices will work on common, open standards, and talk to each other and interact in clever ways. And so, if you walk into the house with someone your security camera doesn’t recognise and your calendar mentions ‘date', some sort of unified learning-based system will dim the lights, turn up the thermostat and start playing Barry White. 

I’m a little skeptical about that. The implied sense of a seamlessly linked and intelligent co-ordinated system isn’t much like the actual internet we have today at all. Rather, it reminds me of Sleeper, or one of those 1960s World Fair ‘house of the future’ mockups in which everything was made by GE or Westinghouse. And it also isn’t what happened with those electric motors or chips. People didn’t buy two dozen electric motors in various shapes and sizes and take them home and build labour-saving devices - they bought fridges and washing machines and blenders and microwave ovens as individual use cases, one by one, over a period of years (or decades). Equally, it seems like the route into IoT is specific services - Nest, Sonos, Fibit, Apple TV/ Chromecast and so on. This is, one might say, a little more like apps than like the web - with Home Depot as the app store. 

Another question is where the value sits - which really means where the ‘smart’ part sits. There’s no one answer here. In some cases it seems pretty clear that the connected thing itself is just the end-point for a siloed cloud service - a diagnostics module in a boiler, or a smart meter, for example. You might even have no actual end-user interaction at all. A Nest is somewhere in between - how much comes from the stand-alone device and how much from the cloud? A Chromecast or Apple TV adds ‘smart’ to a TV, relegating it to dumb glass, but it’s really the smartphone or tablet that has all the intelligence to drive it, with the cloud playing the part of ‘dumb storage'. The same may well happen to cars. Many wearables also feel like they should be satellites for a smartphone (to the extent they're not just subsumed into smartphones anyway), either as a remote sensor or a remote display, but the value again comes from the cloud-based analytics: is it more useful to know how many hours you slept or to get big-data based suggestions as to when you should go to sleep and when you should set your alarm? iBeacon is another fascinating part of this dynamic, because iBeacons themselves are not connected to anything, but again, they add intelligence to the physical world. So every wall or retail display or suitcase or package can become a piece of data. 

That is, sometimes the device is dumb glass (or a dumb sensor), driven by the cloud. And sometime the cloud is dumb storage, driven by the device. 

There’s an interesting Apple/Google dynamic here, of course - if most of these 'things' are some combination of smartphone satellite and cloud end-point, where is the value and control? Apple’s hardware/software integration means it’s best-placed to make things work well (especially with BTLE) but Google is better placed to do much of the cloud stuff. (Of course any device that runs an actual OS will probably use Android, but that doesn’t mean Google will know anything about it). This is the Apple/Google ‘frenemy’ issue all over - if you use Gmail on an iPhone, but only use Gmail and only buy iPhones, both companies win, but both feel insecure. 

Back to the Barry White problem, though: there are clearly interconnections that would make sense, from connecting your Nest to your smart meter to connecting it to your alarm system ("turn the heat down, there’s no-one home”). But if the ‘thing’ is basically a dumb sensor and radio and the clever part is in the cloud then maybe any interconnection should probably be in the cloud too? This is part of the story of IFTTT (an a16z investment) - making ‘recipes’ for connecting discrete services. Or it may be a proprietary BD deal, something like Mint - ‘this services connects to all 26 different smart widgets’. And then there’s the Nest story: put a hub into the home with one clear use case and then build outwards - smoke detectors first, then, say, door locks, alarms, garage doors, lights and so on. 

But equally, there will probably be lots of ‘internet things’ that never speak to each other. A connected TV and a alarm system can probably work just fine without ever knowing about each other. If you install a burglar alarm (for Americans, burglarizerer alarm) today, it will probably use wireless sensors connected over an IEEE standard to a control box that connects over VPN on a mobile data network to the alarm company’s NOC. It will use IP. So lots of proprietary technology has dropped out and the smartphone war dividend has been reaped, but it’s not in any meaningful sense on ‘the internet’ as we normally use the term. So half the ‘internet of things’ devices in your ‘connected home' won’t be on the internet or be connected to anything else.