In this book, I interchangeably compare my ideas about self replicating machines to living things, to economic models, and to information science. I think this is appropriate, because I believe it to be obvious that self-replicating machines will have great impact on, relevance to, use of, and similarity to all of these fields. By looking to all three spheres of thought, and noting their relationships, I believe it possible to come to a better understanding of the future course of things.
Biology is the only field of science where self replication is taken as a given. Life is the only greatly successful system of self replication we can study, and there are billions of years worth of trials and errors to teach us. The theme of emergence is continuous, running from present day living things, all the way back in time to the theoretical beginning of life, where it’s widely held that life emerged from a complex interplay of simple self-reinforcing chemical reactions. In that way, life just started all on it’s own. It wasn’t created or designed, it was an inevitable consequence of a set of preconditions that came together naturally.
Emergence exists in economics as well. Because human behaviour is often planned and deliberate, there is of course managers and designers of systems, but that gives us an incomplete story. Given a bazaar of ideas, it’s not clear which economic policy or social ideology will grow and guide us in the future. All economic things, such as booms, busts, fads, innovations, markets, and competing models and theories of how it all works seem to burst from humanity as if from nothing. Yet there is a sense of how it all works and there are some near universal truths as to how things work, such as the emergent laws of supply and demand that govern so much of our daily lives. Without the profit motive, the machines we use would likely not be continuously refined with such a seemingly singular purpose of efficiency and utility. Our machines are not self-replicating, yet, the concept of evolution of our machines is well known, followed, and even predicted. Much like the complex molecules of life likely evolved in a state of non-equilibrium before they became truly self replicating, our machinery evolves before it can reproduce itself. Human driven evolutionary replication of machines is commonplace.
Even in the more deterministic world of computers and information science, emergence exists. From the humble beginning of Charles Babbage’s proposed Difference Engine, learning how to manipulate ones and zeroes has given us the Internet as a wonder of the world. As with our machines, it didn’t so much get built, as evolved. It is still emerging and evolving at such a rate, that it is only safe to predict basic improvements on existing systems. As a whole, with completely novel systems and inventions constantly appearing, it would be foolish to predict what happens in anything other than the shortest lengths of time. In the complex space where people, the economy, and computers intersect, many kinds of replication happens. From computer viruses and worms roaming free, memes and pictures of cats spreading like wild fire, man-made but automatically installing updates to software, and wholly theoretical beasts emerging from cellular automata, the forms of replication are diverse.
What these systems have in common, is that given the proper conditions and feedback, replicating and self-replicating systems appear almost spontaneously. They grow and evolve following general rules, yet never cease to amaze us with newly discovered complex behaviour. Keeping this theme in mind, I’ve looked at past research efforts on self-replicating machines, and noticed a pattern.
While reading Robert A Frietas Jr. and Ralph C. Merkle’s seminal 2004 survey of the field of self-replication and references therein, it occurred to me that most researchers took the most straight forward approach in their efforts. That is, to attempt to design and/or make self-replicating machines. A scientist’s sensible urge to theoretically model, design, describe, and in some cases, build a prototype self-replicating machine is, after all, the whole point of this field of research. Doing anything else seems irrelevant and unproductive.
I wouldn’t dare call any of these esteemed researchers misguided or mistaken, for they have brought the field to the current state that it is in. However, I have noted that arguably the most successful recent developers in the field have done just that: They de-emphasized the goal of modelling and making self-replicating machines in favour of other goals, and in doing so, have somewhat counter-intuitively, brought us much closer to the practicality of self-replicating machines.
The noteworthy exceptions are lights-out manufacturing, and rapid prototyping. Neither of these have the explicit main goal of making a self-replicating machine, though the idea is certainly part of the thinking. The goals are instead, manufacturing, profit, and improving and revolutionizing existing manufacturing systems. Manufacturers wouldn’t make a lights-out shop unless it padded their bottom line, no matter how cool or partially self-reproducing or repairing it was. Makers of rapid prototyping machines are far more preoccupied with other amazing and useful outputs of their machines, than the fact that this brings us closer to the goal of making machines that make more machines. Even the famous RepRap designers, whose stated goals are to build a RepRap mostly by using parts made by another RepRap, understand that the main use of this machine is all the fun, useful, and profitable things that it can make that aren’t other RepRaps. If partial self-replication was all the RepRap was capable of it, it’s likely the endeavour would have been stillborn. If usefulness and subsequent market penetration and profitability are measures of success, then it appears that emphasizing usefulness more than self-replication is the way to go.
Shifting focus to more than just designing and building fully functional self-replicating machines from scratch leaves us with a different approach:
A piece by piece approach can be taken, where focus is made on just one trait of self-replicating machines, and the research and development associated with it. These kinds of developments would work towards the whole, and are likely have practical uses as well.
Taking the cue from the increasing successes of the fields of lights-out manufacturing and rapid prototyping, employing the concepts learned so far about self-replicating machines can lead to revolutions in many other diverse fields. I believe there are many industries and fields of study that could benefit from being re-thought from the ground up with self-replicating machines in mind, with the goal of increasing usefulness, productivity, and profit. Scientists versed in self-replicating machines may soon find themselves employed in a diverse set of previously unrelated fields which would benefit from varying degrees of self-replication.
A holistic approach also has merit. Examining the relationship self-replicating machines have with computer science, the economy, and biology is still mostly unexplored territory. Much like biological life four billion years before, self-replicating machines are waiting for the right conditions for their emergence. Armed with better knowledge, we can facilitate changes to bring about these necessary conditions. For instance, asking which economic conditions would best lead to the research and development of self-replicating machines could lead to some surprisingly useful economic policies, profitable business ventures, and positive feedback loops.
What these approaches have in common is the presupposition that the fine details of these systems will be worked out in time, and that there is nothing inherently difficult in any specific function of a self-replicating electromechanical machine beyond a good exercise in engineering and a bit of smart design. That being said, the very hard problems lie elsewhere. While solving a small engineering problem is not so difficult, the sheer number and diversity of problems to solve to bring a system to even the most rudimentary levels of self sufficiency is mind-bogglingly large. It is my opinion that no person could design and build more than a tiny subset of the aspects of a self-replicating machine, which is one of the reasons why I advocate a technology management approach over designing a self-replicating machine outright. Another analogy that comes to mind, is an economic one: Designing a self-replicating machine is equivalent to designing a centrally planned economy where every worker is a robot. In my opinion this is an exercise in academic masochism. What I advocate is allowing well known market forces to do a lot of the design, building, and maintenance for us, leaving us as managers of the system as a whole. History has shown us that the efficiency of a managed capitalism, while far from perfect, is far more efficient than the logistical nightmare of a centrally planned economy.
On the other hand, transitioning our economy to one where every worker is replaced by a machine seems like something we have been doing since the beginning of the Industrial Revolution. One of my goals for this book is to attempt to model this transition into the future.