All of the amazing things that you and your peers in this room and around the world have imagined, designed and created, all of the building, bridges, machines, cars, products, and devices... they all have one thing in common. They're all dead.
– Jeff Kowalski, CTO, Autodesk
I'm currently woking on an advanced research effort as a visiting fellow at Autodesk. It is code named Project Primordial. It builds on thinking from Trillions (a book I co-authored in 2012) and in some ways paves the way for the sequel. Trillions was a field guide to the age of the Internet of Things and beyond. Our basic assertion was that:
• We are entering a world saturated with computation.
• The seed of connectivity will hit this supersaturated solution and flip the sock inside out from information "in" computers, to us living "in" the information.
• There are new design and business methods that will be critically important when this happens.
• This shift will lead to an era of unbounded, often malignant, complexity.
As a designer, I'm not interested in doomsday scenarios. It has always been far easier to destroy than to create. I'm far more interested in the question: How will we design and shape this new world to take advantage of the power that it can bring to ourselves, our communities, and to humanity itself? In Project Primordial I'm exploring the intersection of Trillions with two complementary trends—digital manufacturing and machine learning—that dramatically impact this question. I believe these three trends, taken together, give us for the first time in our history, the ability to shift to an entirely new set of design and business paradigms.
These three technological trends are inevitable. They are a done deal and the first signals from the coming deluge are all around us. The challenge is how we surf these waves, what we do about them, and how the act of designing "things" of value will change. Those of us that figure it out sooner rather than later will have an unfair advantage, plain and simple. While others will be reactionary and surprised at each turn of the screw, I hope some of us can not only survive the riptide, but harness its power for good. If you're reading this, my job is like preparing you to gamble in Las Vegas where the odds are not in your favor. You'll have to learn how to count cards, rather than relying on luck and good looks alone.
For this article I'll assume you "get" the basic assertion of Trillions and instead will focus on the role of shaping or authoring products and environments when one of the other trends listed above, namely, machine learning and big data changes the very nature of things.
Let's start with a cup of joe
You are sitting across from me at the café and pick up a cup of steaming hot coffee. Inside my head a collection of neurons that have to do with my own body going through the motions of drinking coffee begin to have a conversation. Will you offer me some coffee? Will you drink that coffee and burn yourself? Will you throw the coffee in my face, grab the table with your other hand, flip it in the air, kick it through the window and jump into the idling car waiting outside (hey, I've seen your Netflix queue)? Or will you blow on the coffee a bit and set the cup back down?
My neurons are mimicking your behavior, inside my head. Scientists call these cells "mirror neurons." They have been observed in primates and humans and in primitive forms in other organisms. Some scientists theorize that mirror neurons help us learn behavior through mimicry, or form a theory of mind about how other people may think and act, help us have empathy for others or even support the creation of our sense of self. They may even help us answer questions like, where do I end and you begin?
While the jury is out on exactly what they do—if they are some sort of special cell type, regular neurons doing double duty, or an artifact of a much deeper and more complex system we haven't discovered yet—the idea of mirror neurons helps us think about how we work, collaborate, and survive in a world where we are surrounded by others that all have a mind of their own, a living essence, that elusive idea of a soul. We spend many of our waking hours trying to decode the nature of the people around us.
Time to go for a ride!
We leave the café and jump into your new self-driving car conveniently idling outside. Since you don't have to watch the road we spin our seats around and continue our conversation. A few minutes later—while we are deep in an argument over the true meaning of life—tires screech, temporary restraints surround us and we hear a crashing noise just outside our window. We look up from our musings and see that your car has narrowly avoided hitting a guy who just ran out into the road to catch a wayward soccer ball. But your car didn't do it all by itself. We were going pretty fast and the car's momentum made it impossible for steering and braking alone to avert a deadly accident. Newton's laws of motion are hard to break. The road surface, two neighboring vehicles, and a streetlight came to the rescue as well. The road roughed up in front of our wheels to increase friction. One of the cars rammed into our side to help make sure we were deflected beyond a margin of error to ensure that we saved the man. The other one slowed down gracefully to save the human-driven cars behind us from creating a pile up. The streetlight changed from red to green and cleared a path for the diverted vehicles to continue on their way.
That scenario doesn't sound much different than the way any well-oiled team plays after years of practice. Whether it's a team of basketball players, doctors, or dance troupes. The collaboration that arises from our ability to mirror and simulate the behavior of others—and build up a body of data for how to react when there is no time but to trust our instincts—is a critical building block in the construction of a vibrant, agile, community. But will everyday things, like cars, roads, and streetlights ever really have something as complex and central to what it means to be human as mirror neurons?
What about when products wake up?
While our car trip sounds like a flight of fancy, consider this: In January Mercedes unveiled a concept car designed for lounging instead of watching the road. A car park in Germany just announced a new robot valet that can park 60% more cars than a human driver in the same space. A recent CNN article about RAY the robot notes, "A new tie-up with Volkswagen announced this month aims to increase the efficiency of RAY by getting the car and the parking robot to communicate with each other." Audi's self-driving car—nicknamed Jack—drove 550 miles from San Francisco to Las Vegas to attend the International Consumer Electronics Show (CES).
Mercedes F015 concept car.
Arguably "Jack" has an early version of mirror neurons. On that trip it had to know where it ended and other cars—and more importantly people—began. It had to run simulations of all the other cars, people, and other things around it to figure out the best path to take on its journey. It had to learn over time how to be a better player in the bigger game of life. Google's self-driving car has already been called out for having to make tough ethical decisions like: Should it let you hit the school bus of kids or drive itself (and unfortunately you) off a cliff? Up to this point those sorts of decisions were left to chance or for Spock and Kirk to figure out.
Another development demonstrated at CES was Qualcomm's "Zeroth" processors. They are the next iteration of a computer chip that has machine learning built in. Zeroth processors were inspired by nature and Qualcomm's goal is to shift from CPUs to NPUs (Neural Processing Units). They demonstrated these chips by embedding them into 3D-printed robotic assemblies that could dynamically detect patterns and shapes. Today we see things that are connected to the network gaining computational power through a sort of infinite computing through the cloud but Zeroth shows that things will have computational power to spare built right into themselves as well.
Thanks to Moore's law, it's safe to say that whatever capabilities a self-driving car has today, your watch, shoes, or headphones will have tomorrow. As an aside, a recent review of Amazon's Echo speaker had this curious note, "I forget she's there sometimes. If I mention her name in passing, she wakes up, a ring of blue twinkling like a mini Northern Lights, and starts telling me about the Gospel of Mark as if she's in the middle of a dream."
As more things get connected—and we begin living in a shared sea of information—these new "lifeforms" will join the flow and find their own environmental niche. We'll have to try to create symbiosis between us and them. Perhaps we should be studying ecological design rather than environmental design. This will be more like growing a garden or raising children rather than like building products, houses, and factories. When machine learning and big data become ever more embedded into our products and environments, the very nature of things will change.
Design for emergence
Designers have always had to consider what limitations and potential values they build into their objects. A product with "good genes" has often had a better chance of surviving in the marketplace. But as things gain mirror neurons (an understanding and empathy for the other things in their environment) and more importantly the ability to change and grow over time, designers will literally become parents of new beings in the world. How will we design for this sort of emergence? How will we shape this "networked matter" at birth to give it (and us) a better chance in the world? Are there new kinds of design tools that could help us plant generative seeds that only blossom if given the right environment to grow? Is there a way that we could build a little intent into our things so they strive to be the best thing they could be, not just at birth but over their entire lifespan? When we plant a real seed for a tree it doesn't give up when it finds the going tough, when its roots hit some rocky soil. It grows around the blockade. When a seed is planted on a hillside the roots dig in at an angle. The tree has genes that get activated at different times so that it is always aspiring to the essence of what it means to be a tree. Could designers build this latent potential into their products when they design them? Could they somehow "ship" their design intent in the product itself?
As with the age-old debate of nature versus nurture, these new machine entities that start out with the best of intentions and with a good set of genes, when set free in the wild, might end up with other plans. Consider a recent art project where the artists gave a software robot a weekly allowance of cash to go on a shopping spree across an area of the web called the Darknet. The bot started out buying mundane things like counterfeit jeans, and Nike trainers, but soon was purchasing baseball caps with hidden cameras, and packets of ecstasy pills. It isn't much of a stretch to imagine the bot realizing it could make a little money on the side, and buy more stuff, not only buying the drugs, but selling some of those pills as well. Pavlov's dog learned to salivate when a bell rang through a process called "dopaminergic learning" or positive reinforcement. It turns out that same process is how you "program" Qualcomm's Zeroth chips. When machines start to learn will we have the emergence of Pavlov's Cog?
Enter a caption (optional)
A sense of place
Aside from the obvious questions—about who or what should be arrested for breaking the law when our machines start to misbehave—this art project highlights another emergent signal about the changing nature of things. Things will become buyers and sellers within the economy. Connected places and things will be a part of the "information carbon cycle," where information is never lost but instead one entities exhaust data is recycled as food to power another's processes. They will be economic players in their own right. Life has always been a market economy where survival of the fittest was the currency of the realm. The way a given organism was shaped, how it was nurtured or challenged by its environment and the other members of its ecosystem, often conferred economic value. Over its lifespan, it gained currency that could be used to barter for things it couldn't do itself.
Winemakers often talk about the "terroir" of a region and the sorts of wines that can only come from a given locale. That sense of place—its unique combination of elements, like climate, geology, geography, the unique life that a given local ecosystem contains—shapes the outcome of a given grape. Grapes from a particular terroir have more economic value—because of the way they were nurtured—than grapes from a different place. What happens when the things we design get nurtured in different ways? Will furniture that is part of the connected ecosystem of all the Airbnb beach houses in the world gain insights that can be sold to others in the economy, particularly those striving for better living spaces? What about clothes in a dormitory or cars that are part of a weekend race club? How will our environments become authors of these new things and how can architects, product designers, and businesses build more virtuous economies?
If we do it right we'll have far more insight and agility than those that have only made it far enough to make dead things or worse yet, zombies that consume our brains in their quest for attention. Just imagine if a thousand other things in in your home were as needy as Amazon's Echo, and as senseless. Our products, designed with life and community in mind could be an integral part of our design team. To paraphrase and extend Bill Joy's Law, there are always more smart people—and products and places—outside your company than inside. How can we harness the value of all those things and places, all that nature and nurture as this new era comes into focus?
Amazon's Echo, a hands-free, voice-activated connected speaker.
When products become social
In Sandy Pentland's recent book, Social Physics, he reports on groundbreaking research into the ways that ideas flow within human social networks. In his research he uses a vast number of "digital breadcrumbs" (see Trillions above for why that's becoming easier than ever before) that stream in from various connected things to track how we—as a group—coordinate, discover, engage with, and exploit ideas. His team's experiments, and the underlying "physics" that he has discovered, can predict how effective or intelligent a given social network is. The underlying math they've developed is a down payment on discovering ways to improve the collective intelligence and performance that a group of people is capable of. He shows that the value of a sensor embedded into your new product or environment is not very high until it can be added to the social network of other things and people that care about that flow of data. Often the biggest value comes from not only collecting that sensor's data over the "arrow of time" but also combining it with entirely different data sets. But what will happen when products and environments enter the social network? When they shift from passive, dead, things to players in the world of social physics? How will we apply the algorithms of idea flow to things (and places) that collaborate with people? Could we build design tools that help us simulate and tune the physics of a particular place or group of people and things to radically improve the group's performance along a given dimension?
A scene from RPG, Shadow of Mordor.
As we talk about social interactions it reminds me of a recent development in game design and makes me think that designers of things and places could learn from designers of characters, games, and dungeons. Building enchanted, delightful, surprising, or lovable products takes on a whole new meaning when they have the ability to become actors in our shared drama. A game system called "Nemesis," an integral part of The Shadow of Mordor, shows what happens when things (in this case artificially created characters) hold a grudge. The Nemesis system is a major innovation in game design. If you fight a band of Orcs and some of them survive, they share their experiences with others as the game plays out and remember that you hurt them. If you ever run into one of the Orcs again, you discover it's been holding a grudge. You also find out, sometimes the hard way, that it has grown and matured and has more skills than it did before. Players of the game find this new mechanic as enticing as catnip. Some—possibly simple—social interactions make the game seem far richer and more vibrant than the mechanistic cardboard cutout characters of the past. It seems that the other "entities" in Mordor's game have a life of their own when you aren't looking at them. They have a living essence, their own particular nature.
We believed that an interactive story doesn't mean letting players choose between a set of branching narratives that we have created; it should mean that players are at the centre of authoring the experience and that the world of the game remembers and responds to them. – Michael de Plater, Mordor Design Director
There are two key words in Michael's quote that I want to draw attention to: "authoring" and "remembers." Products and environments have always had that 5% or 10% subset of users that modify, customize, or tune things to fit "just right." They are para-programmers, scriptors, ultimately participants in the authoring of the product, just not at birth. They sometimes fix bad products and other times extend good ones to be better. Products have "remembered" too, but only through wear and tear and the general shifting that a product goes through as it is used throughout its life. A new car wears in a bit and becomes that perfect summer ride with a personality all its own; a baseball glove becomes an extension of a ballplayer's hand over time. But consider a world where things have machine learning, sensing, and connectivity built in.
Could we, instead of letting entropy dictate the memory of things, reverse entropy and practice our products into better shape? What sort of embodied potential could we design into our things so that our users become authors and maybe even participants in the economy as well? How could living things shift the environmental picture? What would happen if things had a natural lifespan and could contribute to the local ecosystem when their task was done? The concept of cradle to cradle takes on a much deeper and more profound meaning when our things can be a part of our social network. How could we use the feedback loop that connectivity provides to flow resources and ideas back to other things within the community? To enlist more travelers on our quest or help the producers and consumers of those things make the next generation even better? And, like dungeon masters, how could we design vibrancy into our places and things as a response to new challenges, not only at birth, but while the game is playing out?
"Simple things should be simple, complex things should be possible." – Alan Kay
In Trillions we noted that we should try to follow nature's lead and make complex things out of simple things. For instance, atoms make molecules, molecules make cells, cells make organs, systems, you, me, and us. This idea of "layered semantics" is critical to the design of complex systems, yet it may seem at odds with my comments about nearly every "thing" having self-driving car level intelligence. Another thread of design science research, this one going on at my home lab at MAYA (under the name "Interstacks") is designed to address just that tension. In that work we are exploring ways of making complex systems—built out of atoms and bits—tractable, scalable (to galaxies not just planet Earth), and comprehensible. While on the outside the component architecture of Interstacks looks at first blush like any other DIY electronics kit, it has been designed to address both the runaway danger of putting too much computation and complexity too far down in the layers of your things, and also to allow for the casual development of little things like connected products, bigger things like connected houses, and extra large things like stadiums and communities with the same set of tools.
We have three mantras in that research theme:
• The least amount of complexity to do any one particular job, rather than the most.
• Develop open components that are fungible between software and hardware.
• Prototypes not PowerPoints.
Each of them work together to play a little trick on the world. A sleight of hand that pledges something ordinary like a maker kit, but turns it into something extraordinary (like living products and environments that can join a community), and in the process brings it back (the prestige) to a level of layered complexity that makes it possible for us to scale without bounds and tame the noise down to a level where we can thrive. If we can foster a community of makers that can understand what they've created and the implications of the complexity they intend to bring into the world, and we can offer them a way to use recyclable bits and atoms in a market economy, we can tame complexity and move it to the right level, much like large marbles shuffle to the top of a jar full of sand when agitated.
Ultimately, if normal people can become the parents of this new breed of life in the world (just like all of us regular Joes seem to be able to do with our existing children), then we won't become trapped by the technological elite who at times enjoy the hobbies of complexity and get distracted away from the unmet needs of our customers and communities.
For an example of the dangers of this sort of hobby, consider this question…
Humans are fooled by randomness. We have a hard time determining between coincidence and causation. We believe lots of things that are crazy, sometimes because it "feels right" and other times because we have incomplete information but want to believe we can understand the world (or have to act even if we don't have all the information). We are pattern finders by nature even when there aren't any patterns there. Evolution may have encouraged this sort of behavior. The downside of imagining that something going bump in the night might be a bad thing trying to eat us was that, if it wasn't a predator, we just looked foolish (but were still alive). Science came along at some point as an antidote to this sort of thinking. But science turns out to be hard in practice. A seminal paper by Pedro Domingos about the "black art" of machine learning notes that machines can also be fooled. They can have a hard time determining between coincidence and causation, or learning how to generalize a new skill from specific example data sets. We are in the early days of learning how to design machines that learn. I don't think we can make them logical right away—not only do we not know how to do that, they (for awhile) have to live in a human economy and humans are not logical—especially in their wants. But hopefully we can avoid affecting our mind-children with the psychoses that we have.
Richard Feynman gave a commencement speech at Caltech that is a master class in the difference between scientific and magical thinking. He illustrates his point by sharing the history of "Cargo Cults" where South Sea islanders—who had a sudden surfeit of riches drop from the sky in the form of cargo containers during World War II—formed a sort of cult when the wealth disappeared (the war ended). They built bamboo landing towers with air traffic controllers inside replete with coconut headphones. Unfortunately, while they did everything right from their point of view, the cargo didn't return. They didn't have a valid model of the actual system (for instance the real model of the world included two epic cultures battling for dominance and the islanders turned out to be a convenient way station). Worse, some cargo cults were formed because a charismatic leader that told of having had a "myth-dream" about what the gods of cargo needed. The mysticism provided a way to control the masses and co-opt the community into doing what the leader wanted.
So what happens if our products form cargo cults? What if they fool our business leaders into making decisions based on flights of fancy rather than facts on the ground? What about when our machines begin to see patterns where there aren't any, or worse, when some of the objects and environments in our social network try to become shamans through their influence and myth-dreams? While they are learning to study us, in the best of worlds to serve us better, what if they begin to foster magical thinking?
Products that begin to act wacky and guided by voices could lose trust among their peers. Designers and organizations that produce those sorts of products or environments could become ostracized or dismissed the same way we roll our eyes when our loony uncle walks into the room. Getting smart about statistics, about modeling, forecasting, maybe even having our products gain literacy in human-centered design science, could be an antidote. But, Feynman notes that the way science works best is when we do something more. It works best when we are brutally honest, capture all of our results even if they don't fit our current model, and become radically transparent about what works and doesn't. Most importantly of all it works best when others can test out our ideas themselves and contribute to the information commons.
For business models the implications—where a given thing plays in the economy, and how much we share versus keep to ourselves— are vast. Machine learning is one way we'll tame all the complexity in our world. It is a powerful spice in the recipe we're cooking up. The sooner we as designers learn to think ecologically the sooner we'll discover the right mixture of nature versus nurture that can help our work thrive. Ultimately my goal is to stir the pot, and see what bubbles up as the necessary and sufficient primordial soup for this new species of product design.
In this article I've posed more questions than answers. It won't surprise you to learn I've got a few initial answers or at least hypotheses I'm interested in testing with a few brave souls. If you're interested in the research we are doing into shaping networked matter, join the conversation or become a partner in our research experiments.
How will designers create, not an Internet of Things (or worse, an Internet of Thing) but rather, as Jeff Kowalski notes in his keynote at Autodesk University, a Community of Things?