Editor’s note: This is part two of a two-part series reflecting on the O’Reilly Solid Conference from the perspective of a data scientist. Normally we wouldn’t publish takeaways from an event held nearly two months ago, but these insights were so good we thought they needed to be shared.
In mid-May, I was at Solid, O’Reilly’s new conference on the convergence of hardware and software. In Part one of this series, I talked about the falling cost of bringing a hardware start-up to market, about the trends leading to that drop, and a few thoughts on how that relates to the role of a data scientist.
I mentioned two phrases that I’ve heard Jon Bruner say, in one form or another. The first, “merging of hardware and software,” was covered in the last piece. The other is the “exchange between the virtual and actual.” I also mentioned that I think the material future of physical stuff is up for grabs. What does that mean, and how do those two sentiments tie together?
In the first sense, “exchange between the virtual and actual” isn’t that different from “merging of hardware and software.” Clearly, when we move more of the intelligence of our machines into software, some of the actual is becoming virtual. There’s that, but there is a deeper trend that came up repeatedly at Solid that I’d like to flesh out a bit more.
My ability to articulate these trends is even less sophisticated, in part because most of these ideas came from future-focused keynotes and some follow-up reading. I am even more out of my depth here, so apologies if I butcher anything on either the manufacturing or the biology I’m about to discuss.
What is our world made of?
The materials of today include the likes of plastic, particleboard, glue, copper, and silicon. At the Ft. Mason center where Solid was held, it was clear which things were from our current age (the laminate signs, the plastic convention tables, the particleboard bookshelves). Had someone set down a chrome and steel appliance in the middle of the convention floor, it would have been just as obvious that it was from the first half of the last century.
Our age is an age of CNC (computer numerical control). Our factories mill, cut, and injection mold to make our personal possessions. Almost everything made today is cut away by computer or melted into shape in something else that was cut away by computer. The prior age was an age of bent metal and cast steel. The age before that was an age of cast iron and wood. Of course, we continue to make things in older ways, but that doesn’t detract from the fact that each age has characteristic materials and manufacturing practices that define its look and feel.
What will the future look like? What will be the default manufacturing style in 50 or a 100 years? I’m speaking more here about small possessions, not buildings or infrastructure, which naturally change more slowly.
The consensus seems to be that the future will be pixelated. Everything today is made of unique components that are cut, melted, or shaped into just the right shape and stuck together. A chair is made of a few curved pieces of particleboard glued together, a smoke detector is made of a few injection molded pieces of plastic screwed together, a lamp is made basically the same way — topologically smooth, continuous parts, joined roughly.
By contrast, in a pixelated world, our things will be made out of reusable, identical parts, the same way that our computer images are made of reusable pixels that can display any number of pictures. What might that mean?
Gershenfeld calls this “digital” manufacturing, and not because the machine that makes the thing is digitally controlled, which is the current meaning of digital manufacturing. Instead, the manufacturing is digital because the component pieces are discrete; made of distinct pieces; and, like a digital signal, either connected or not connected — on or off, with no free-form interfaces, more like Legos than brick and mortar. And just as digital signals permit better copies and lower errors compared to analog ones, so do, in theory, digital things.
I think I have seen the future, and it is reusable, composable, and rapid from top to bottom.
What do I mean? Let’s take an example. Before the advent of digital telephony, call quality degraded over distance. If a person in New York called a friend in Ohio, the quality of the audio would be better than if someone in New York called a friend in California. When Bell switched (no pun intended) to digital representations of calls, it didn’t matter whether you were in Ohio or California because digital signals can be copied with far fewer errors in them than analog ones can.
We take for granted that if you copy a file and then take a copy of that copy, the second copy will be identical to the first. We expect little to no corruption in our data from repeated copying, even after transmission over phone lines and satellite links. By contrast, when you photocopy a photocopy, or make a tape of a tape, quality rapidly degrades. These are analog processes. At the moment, the same thing is true of most of our stuff.
By contrast, to use Gershenfeld’s favorite example, consider a Lego castle. A Lego castle, made from Lego bricks, can be assembled to be exactly the same every time. It can be made, then broken apart back into its pieces, then remade again, without any flaws. The child who assembles the Lego castle doesn’t even need very fine motor control to get it exactly right. All of the intense manufacturing can happen once, at the level of the brick, and after that, the final result can be made (and repaired) very quickly.
There are already examples of this, from digitally made digital circuits to digital polyhedra for arbitrary shapes. We may not be far from being able to build a toy car on a desktop that drives off of the construction bed. The future of desktop manufacturing may be more pick-and-place than plastic squirt gun.
Obviously, it’s not digital all the way down, just as computers ultimately at their smallest level are continuous wires, doped silicon, and plastic parts. We will still need to manufacture the pieces, and the pieces will require large factories and big expensive shipping containers, at least for awhile. But having a layer where things switch from analog to digital can clearly bring huge benefits.
So far we’ve looked at things made digitally at the scales of millimeters and centimeters. What about much smaller components?
The point was made several times at Solid that reusable, composable, repairable digital manufacturing is how biological manufacturing is done. We have small, reusable pieces (amino acids) that are combined together to make objects with highly complex properties that can then be recycled to make new objects. When leaves fall off trees in autumn, those leaves are digested back into the same materials that new leaves will be made from, with no loss other than the energy it takes to remake them. Try doing that with Ikea chairs.
Living objects are much more similar to an LED TV than to an old-style cathode ray tube: small component parts with a large number of discrete states are organized hierarchically, with a lot of repetition to give the desired result. By contrast, with a cathode ray tube, the only guarantee that the light ends up in the right part of the TV is the precision of the electron gun.
This kind of digital manufacturing is also on the horizon. DIY bioengineering is now a thing. Carl Bass, CEO of Autodesk, showed off designs in the latest version of Autodesk 360 for tiny baskets, structurally made of DNA, that release a payload on contact with a protein. Though the manufacturing hasn’t been able to make it as far as the design has, the immediate potential is still huge.
There are other ways the world may become more pixelated. Hiroshi Ishii discussed the possibility of reactive surfaces, able to come up and meet us rather than being pictures under glass. In his vision, our current things are just too darned still, unable to bend or change themselves in any useful way beyond displaying pictures on their surfaces. When our things are made of voxels and tightly coupled actuators, they may cease to be so dormant.
Exchange between actual and virtual
It would seem, then, that we’re only a generation out from something like a replicator, albeit far from the Star Trek kind, able to produce objects with built-in sensors and electronics, constant connectivity, and reactive surfaces. At that point, the distinction between hardware and software basically collapses. When most objects are software-built, highly networked, and full of ubiquitous, built-in, cheap computing, we are really in a science fiction scenario. I may live to see it happen during my career.
Nanotech fabricators promised to us in the 1980s turned out to be premature. We were likely taking on too much, too fast. Getting good at programming matter on a large scale is probably a prerequisite to getting good at doing it on a small one. I think I have seen the future, and it is reusable, composable, and rapid from top to bottom.
What does all of this mean for data science? I don’t know. I think it’s possible that even the people who are promising an explosion of data as we make our toasters and aspirin tablets smart are vastly underestimating just how cheap and ubiquitous data collection and automated decision making will become. I’m definitely looking forward to being along for the ride.