Software Above the Level of a Single Device

Software Above the Level of a Single Device

Tim O’ReillySolid

May 22, 2014

1Thursday, May 22, 14

@timoreilly #OReillySolid

2

“The skill of writing is to create a context in which other people can think.”

Edwin Schlossberg


Edwin Schlossberg once said...That’s another way of saying that the words we use are a kind of map that guides our thinking. And as you know, a good map leads us to our destination, while a bad map leads us astray.



So let me start with a word cloud of the words in the Internet of Things entry on Wikipedia. (By the way, that entry could really use somework, and this group here is probably the right one to fix it.) You’ll see a huge focus on Things and Devices and Objects. And those things are “smart.”



And to be sure, there are amazing smart things here at Solid. I want that Taktia smart router that can make it easy for me to cut straight lines.



And man, that one wheel motorized skateboard that you control with your weight.



And the Makani wind turbine - that uses smart controls to completely rethink how to generate wind power. By the way, that was inventedby my son-in-law Saul Griffith, now at @Otherlab.



I want to highlight two other words in the word cloud. Look at this little word “people.” People are a really importantpart of this emerging future. And I also want to highlight this tiny word over here: “Time.” People and Time are avery big part of the IoT story. We need to get these a lot bigger in that word cloud!


“The Internet of Smart Things”

Sensor and data driven Autonomous No human input Real time

What We Tend to Imagine


Here are some of the ideas that I think are commonly associated with the notion “smart things.” But are they right?Are they a good map? Or do they mislead us?


Which one of these things is smart?

Smart Dumb?


Most of you who have heard me talk about this subject before know that I’m fascinated by the connection between the Googleautonomous vehicle and the Google Street view car. Is the autonomous vehicle smart, and the street view car dumb? I don’t think so. The two are part of a system, that includes both machines and humans. Those humans are augmented by sensors,but are a critical part of the system. And the human input into the system is NOT real time, but is instead time-delayed. Thereare a huge number of key concepts to unpack here.


Human input is critical That human input may be time delayed, as a

learned memory That human input may be implicit rather than

explicit



11Thursday, May 22, 14Or consider the Nest Learning Thermostat. The Nest “learns.” But who teaches it? We do, both by explicit actions, like turning the heat up or down


12Thursday, May 22, 14and by implicit actions, like being in the house or not for an extended period of time.


13Thursday, May 22, 14And of course, if you have one of these things, you realize that a lot of the input you give is the “old fashioned way”, through your phone or tablet.


Sensors allow us to create new kinds of user interface



But remember, they are stillUSER INTERFACES

15Thursday, May 22, 14Explore, test, measure, just like you do with any other UI in the modern world. There is a huge period of experimentation ahead on the UI front. Do wetell our devices what we want by touching them, by talking to them, or just by being around them? How do these modes relate? For example, itseems to me that when I set an explicit schedule for my Nest, it stops learning. (I guess people do that too, which is why it’s best to set a visionrather than be too explicit!)


Beautiful, smart, but didn’t think about how I might want to use it

16Thursday, May 22, 14Here’s a great example of a smart thing that misses the mark because it didn’t think enough about the user. The Tesla key is beautiful. It wakes up my car automatically, and has hidden controls I can double click. But guess what: I can’t put it on a key ring. That’s a design failure.


17Thursday, May 22, 14There’s this great sticker that gets handed out by the UK Government Digital Service - I got this sheet of them from the UK CTO, LiamMaxwell - that says “What is the user need?” The GDS has done a great job of focusing on designing with user needs put first, andthere’s a lesson there for hardware designers too. (GDS Design Principles: https://www.gov.uk/design-principles )



Also, let’s not forget the “Internet” part of the Internet of Things.


19Thursday, May 22, 14It’s important not to look at a device like the Nest in isolation. It’s connected to the phone, they’ve introduced their smoke detector that is networked through the house, and uses the wifi-connected thermostat as its hub....


20Thursday, May 22, 14That’s also true of the Google self-driving car. It’s in a network that relies on GPS satellites, data in data centers. And eventually connections with other cars and smart road infrastructure like traffic lights that will broadcast in advance how long till they are about to change, so cars can optimize their speed.


21Thursday, May 22, 14And within that cloud data center, there are cooling systems that are packed with sensors to automatically regulate the temperature and control the flow of warm and cold fluids through the data center’s circulatory system. And those cloud centers are themselvesnetworked.


“The Network of Networks”

Interoperability! Interoperability! Interoperability!

22Thursday, May 22, 14But if we are going to be thinking about this “Internet of Things” as part of the Internet, let’s make it a real internet. And that means Interoperability, Interoperability, Interoperability. They used to call the Internet “the network of networks” because all the old networksdidn’t talk well to each other. Let’s not repeat that mistake. Right now, we don’t really have an Internet of Things. We have a bunchof Intranets of Things.


The Robustness Principle“Be conservative in what

you send, be liberal in what you accept from

others.” RFC 761

Jon Postel

23Thursday, May 22, 14One of the key tenets of the early internet design was something called “The Robustness Principle.” The idea was that the internet wouldbe robust if people followed what really amounts to the Golden Rule applied to networks. “Be conservative...”

This needs to be true of the Internet of Things as well.


24

“Useful software written above the level of the single device will command high margins for a long time to come.”

-Dave Stutz, “On Leaving Microsoft” February 11, 2003

Dave Stutz


http://www.synthesist.net/writing/onleavingms.htmlThere’s another framing concept about networks that is a key part of my thinking. Dave Stutz, in his parting letterto Microsoft back in 2003, had this great line “Software above the level of a single device.” It has shaped my thinkingever since.

25Thursday, May 22, 14Consider Uber - not generally thought of as an IoT startup. (Though it should be. It depends on the sensors in the smartestdevices any of us carry around today - our smartphones. It’s not just an app. It’s a network. It only works because not only the passenger but a network of drivers are reporting their locations at all times. The whole thing depends on what I’ve long beencalling “the Internet operating system.” You have features like communications, payment, reputation, many of which come fromoutside vendors via API.


“Uber is a $3.5 billion lesson in building for how the world *should* work instead of optimizing for how the

world *does* work” - Aaron Levie of Box.net


But here’s the point I really want to make here: Aaron Levie of Box said it perfectly in a tweet: “Uber is a $3.5 billion lesson in building for how the world *should* work instead of optimizing for how the world *does* work.”

Once “things” become smart, you can reinvent the whole business process, what IBM calls “systems of interaction.” Uber is reinventing the taxi business. What else could be completely reinvented?


27Thursday, May 22, 14Another service that uses the location sensor to rethink a real world process into “how it ought to be” instead of how it is, is @coverpay.This is one of the companies we invested in at OATV. It gives an uber-like experience for restaurants. When you’re done, you justget up and walk out. Your credit card is automatically charged, because the restaurant knows who you are. GPS, images, but alsocredit cards bypassed by cloud registration - remember that mag stripe is a primitive way of getting data in and out via sensor too! There’s a design pattern here: when you know several things about someone (e.g. restaurant reservation + location + face) you can do things (with permission) like charge a credit card automatically. By the way, Cover just launched in San Francisco, and ifyou sign up and use my twitter handle (@timoreilly) as a referrer code, you’ll get a $10 discount on your first order. Drinks on me! (Worksin NY too.)


28Thursday, May 22, 14Another amazing one is Makespace. What they realized is that if you can take pictures, so you know what you put in your storage boxes, they can be stored and retrieved in a whole new way - much more like a modern warehouse retrieves stuff rather than your crowded storage locker loaded with boxes whose contents are soon unknowable. And this illustrates that it doesn’t have to be super complex sensing. It’s more a matterof thinking how to deploy sensors and then ask yourself what you can do differently now that you know something via that sensor.

And remember that the camera and the microphone are among the most important sensors. We are used to them, and we take them for granted.


Work on what is hard

29Thursday, May 22, 14I want to close with an appeal. We are at a moment when we are gaining great new capabilities, and can build applications that would haveseemed magical to people only a few years ago. And it’s still only the beginning!

But don’t just think about building cool new consumer applications. Work on what is hard. The physical world isn’t just something we do forentertainment or information sharing. It’s the place where we actually live, where “matters of life and death” isn’t a metaphorical saying.Think about meaningful real world problems that you can make a difference with. Thank you very much.

Software Above the Level of a Single Device

Technology

timoreilly

word cloud

robustness

dave stutz

aaron levie

human input

5 billion

single device