Section 1.0 - Introduction
Human-defined systems have characteristics of interest, such as size, monetary value, or output rate. A change in these parameters in a desired direction is considered an improvement. Self-improvement is when the change is produced by internal action of the system. Self-improvement has occurred throughout history in many ways, such the as evolution of life, or the development of civilization. But these changes have often been random rather than purposeful.
In this book we consider self-improvement generally, but it is mainly about a particular type of self-improving production system called a Seed Factory. This is a starter set of equipment intentionally designed to improve by a recursive process:
Part of the factory's capacity is used to make more or better equipment for the factory itself. The rest makes useful products, like any other factory. The fraction used for self-improvement can vary over time, according to whatever goals the factory operators choose. The added equipment increases the ability to further improve. Output is proportional to how much equipment you have. So if a significant fraction is devoted to improvement then total capacity can grow exponentially. This idea matters because we face a number of large-scale problems which can only be addressed by large-scale solutions. Exponentially growing systems can provide the necessary scale.
This section (1.0) introduces some of the problems, how the systems engineering method can be used to develop solutions, and some examples of self-improving designs we will detail later. Section 2 goes through the history of self-improvement in nature and in human civilization, and how those led to the problems we face. We then look at self-improvement as an idea. It is closely related to other ideas like progress and economic development. More recently, ideas like self-replicating systems, and technologies like computers and automation, have led to the concept of a seed factory as an instance of self-improvement. It is still a fairly new concept, so section 2 concludes with what we don't know about it yet, and areas for further research.
Section 3 details the current state of the seed factory concept, and the pieces which make it up. This includes a reference architecture for developing system designs. Section 4 discusses the design process. A self-expanding factory is complicated, so we apply the "systems engineering" approach, which was developed to handle complex projects. The state of the factory at any point in its growth will include some subset of production processes and equipment types. These are drawn from the set of all known processes and equipment. We provide a reference list of many of the available options as inputs to the design process.
Sections 5 through 8 provide a series of design examples for use on Earth. The series is linked by the fact that one example can grow into later ones, or produce starter sets for them. The last section of this volume is a set of notes that are too new or don't fit into the main discussion. It also serves as part of our project database.
Volume II of this set begins with the relevant science and engineering methods that apply to space projects. We then continue the series of design examples for various space locations. The environments, resources, and distances are quite different as compared to Earth, so we also include explanations of these differences, and how the varying conditions affect the system designs.
Through history, large-scale systems like cities have grown, and made major improvements to themselves. But intentional design for growth and self-improvement has been rare. Designing from the start to avoid side effects like environmental damage and pollution is even rarer. So these volumes are mostly future-oriented, describing types of projects that haven't been done before. This is in comparison to mature fields like Civil Engineering, where there are many past examples of construction projects, and many people with skills and experience in developing them.
1.1 - Some Problems Facing Civilization
These are engineering-oriented books. Engineering in general involves applying knowledge towards useful ends, such as satisfying people's needs, solving problems, or making life better. If seed factories are to be more than just an interesting idea, they should be developed with such useful purposes in mind. So we begin their investigation by identifying some current and future problems facing civilization. This is by no means a complete list. Rather, they are problems we think can be addressed by technical solutions. The ideas and methods in these books can then be applied to develop proposed solutions.
- Insufficient Development
Many places where people live have less of the artifacts and systems of civilization, and the benefits they bring, relative to highly developed areas like large high-income cities. Even in the most developed areas it would be difficult to find people perfectly satisfied with their current situation. They can usually list a number of things they would like to improve, if only they had more time, money, or other resources. There is no shortage of desire for improvements. What is lacking are the means to make them happen. This is particularly true in difficult or hostile environments, where the extra work required has left them nearly or entirely undeveloped.
In addition to uneven development, even the most developed areas have an uneven distribution of resources. Those who have the least lead a precarious existence, where any adverse event can seriously degrade their lives. So in addition to general development, we would like to provide backup systems, reserves, and other buffers against adversity.
- A Deteriorating Environment
A healthy environment is necessary for us to live on Earth. But side effects of our current civilization have led to changes for the worse. Excess carbon dioxide (CO2) leading to climate change is probably the most pressing environmental problem. Others include deforestation, desertification, species extinction, and various kinds of pollution, like airborne particulates and microplastics.
- Technological Unemployment
Recent centuries have seen accelerating replacement of manual labor with powered equipment. This produced a shift from farming, which used to be people's main activity, to manufacturing and services. Continued improvements have reduced manufacturing labor, to where services are now the dominant activity in developed areas. Since about the mid-20th century, technologies like automation, robotics, software, and artificial intelligence have enabled "Smart Tools". These have an increasing ability to work on their own, replacing much of the remaining labor.
Our modern economic system is based on people trading their specialized labor for money, then the money for the other things they need and want. If most labor can be replaced by smarter tools, we get the problem of technological unemployment. The unemployed can no longer buy what they need, and the still-employed lose them as customers. The system of trading work for money and then money for other goods breaks down.
This is not yet a severe problem. Through history, people displaced from one kind of work have found something else to do. But the recent advances in smart tools threaten to displace many jobs in fields like transportation, distribution, and retail in a short period of time. The displaced workers may not be able to find new jobs in other fields. It is worth thinking about how to deal with this potential problem before it becomes severe and we are faced with mass unemployment.
1.2 - How to Tackle These Problems?
Our civilization, and the parts that make it up, are very complex systems. A method called Systems Engineering was developed in the 20th century to deal with complex systems. We propose to use it to address the problems noted above. The systems approach starts by defining the unmet needs which one or more systems are intended to satisfy. For existing systems this is in terms of what changes are needed to them. Identifying what is wanted comes before any consideration of designs or solutions. Having the answer in mind before stating the problem prematurely limits considering all the options. Goals should also not be limited by what is thought to be possible now. Partly meeting the goals, and identifying what research or improvements are needed to meet the rest is a better approach than limiting our goals and going no further.
We identified some problems in a general way at 1.1. For the systems approach we need to specify what we want more exactly. That way we can measure how well proposed designs or solutions meet the goals, which ones are better or worse, and in what direction to make improvements. We will start that process here, and continue it in later sections. Since civilization already exists, these statements are in terms of changes. The goal statements (in bold face) are followed by some explanations.
- (1) Bring less-developed parts of civilization up to [TBD] level of satisfaction with their situation, given sufficient information about their options. Increase satisfaction in more developed areas by [TBD] and provide means to deal with adverse events. - TBD means "to be determined". We haven't defined how to measure satisfaction levels yet. We know we want some kind of measure for it, so we use [TBD] as a marker for later work to fill in. Significant numbers of people lack education and information on what is possible, and therefore can't know what they would want if they had that knowledge. Supplying them with that knowledge so they can choose is part of the goal. Satisfaction with the current state of things is not enough if something bad happens, so we also want to be able to handle adversity.
- (2) Reverse human-caused environmental problems by [TBD] percent while minimizing side effects. To the extent possible, also minimize natural risks. - All human activities have side effects, so we can never entirely eliminate them. The goal is then to minimize them, and make them less harmful. The natural world also presents many hazards, like hurricanes and earthquakes. We can't eliminate these hazards either, but we can minimize the damage they cause.
- (3) Avoid or minimize future problems, such as technological unemployment. - Permanently displaced workers due to automation is one problem we already identified. There may be others we want to deal with ahead of time. Part of the goal is to identify them.
1.3 - Systems and the Systems Engineering Method
A "System" is any collection of elements which interact with each other and the outside world, and performs some identifiable function or transformation. For example, a computer system allows you to read this document on Wikibooks. It contains elements like a power supply, semiconductor chips, and software that work together. It interacts with you as the reader, and the Internet to access this page. If you were editing this page, you would be transforming it from one version to the next.
We define systems to help with analysis and design, and to help meet intended goals. Every system has a boundary. This is a mental construct rather than a wall or a fence. It separates the system from what is outside the system. Sometimes system boundaries are natural and obvious, such as an airplane vs everything outside the airplane. Other times system boundaries and their contents are arbitrary and in multiple pieces and places. Defining them can still be helpful in working on a project.
A wide variety of things can cross the system boundary, and thus enter or leave the system. For example, a factory system can take in raw materials and power as inputs, and deliver finished products and various kinds of wastes as outputs. Physics tells us matter and energy can't spontaneously appear or disappear, only change form. So mathematically, the change in the contents of a system equals the flows entering the system minus the flows leaving.
Systems can be any shape or size, and can contain other, smaller, systems. Smaller systems within a larger one are called "subsystems", which can in turn contain even smaller subsystems, to any level of depth. The human mind can't comprehend the entirety of a large complex system. So a key method in systems engineering is to break up large systems into smaller subsystems and sub-sub-systems, until they are small enough and simple enough to understand, design, build, etc. The flows at every level must still equal the change in subsystem contents, so straightforward bookkeeping can ensure nothing has been forgotten or lost in the total system.
Now that we've identified problems and set goals, the next step is starting to identify potential solutions that can meet them. At the start, we don't know what the best approach will be. So typically multiple candidates will be proposed or formulated. Over time, most of these will be weeded out and optimized, until one or a few best options remain. Research and testing may be needed before the final choices are made. The details of the process are more fully covered in Section 4.
The premise of these books is that seed factories and self-improving systems can be part of solutions to the problems noted above. That needs to be proven, though, and not just assumed. The examples we present and analyze will help to do that, but they need to be put to actual use and shown to work as expected. Self-improvement is a design feature. It is not a complete solution by itself. System designs must also include and build on past experience in fields like Industrial Technology, other Engineering fields, and other parts of our accumulated human knowledge. Those other subjects will be referenced and included where appropriate.
1.4 - Design Examples
Sections 5 through 8 describe four types of self-improving production systems. Each is intended to help solve the problems and meet the goals noted above. How we do this is by using particular approaches and features in their design.
The first problem is that of insufficient development in terms of modern artifacts and systems. In principle this can be resolved by producing and delivering the needed items. But that presents several secondary problems. People in low-income areas may lack funds to pay for such things. Even if they were donated, they may not have the skills and experience to operate and maintain them. A solution to both secondary problems is to bootstrap their own production from seed factory starter sets and their own labor. They will need less funds to build the factories if they self-build them, and they can then make the other finished products they want. The experience gained in building the factories and making the products will then enable them to operate and maintain them later.
In more developed areas, there are still widespread shortages of time, money, or resources to make improvements and increase satisfaction. Our approach to this is self-improving systems with a high level of smart tools. Starter sets take less money, smart tools take less time to operate, and the production systems can produce their own energy and obtain material resources. Due to the higher level of complexity and skills needed, these systems would likely be developed as cooperative projects involving many people.
The second problem is risk from environmental damage and natural events. A start towards solving it is education about these risks and what can be done to reduce them. New systems can be designed to incorporate renewable energy, avoid or recycle wastes, reduce pollution, and maintain natural habitats. These design features would be included in both finished products and the production systems that make them.
The third item is addressing future problems before they become urgent. An early task is to identify what those problems might be, besides technological unemployment that we already identified. Then the systems approach would be used to identify solutions or preventive measures, and incorporate them into system designs. The unemployment problem could be addressed by production systems of the types in our examples. As owners they receive a share of their output directly, regardless of how automated the systems are, or how much labor is needed for their operation.
The four examples form a linked series of projects to build. One way to follow the series is by evolution. A large enough series of improvements and growth can turn one example into the next. Another way is by setting up new seed factories for a later example. Ideally, the various parts for new seed factories would be made by previous ones. A later example could also be started fresh by supplying equipment from elsewhere, or by some mixture of internal and outside sources. However they are started, the experience from earlier projects and their growth paths can be applied to later ones.
The example systems are:
- Personal Production (Section 5)
This example has the goal of making a range of basic products for a local community of owner/operators, such as food, building materials, and utilities. It is personal in the sense they make things for their own use and others in their community. It is not like a personal computer in having just one owner. It begins with a network of people interested in working together and conventional workshop tools. At first, production is at hobby and home-use scale. This keeps early costs within reach. To this they gradually add a starter set of seed machines. The combined collection is partly used to make more equipment, alongside the basic products.
Later equipment is scaled in size to increase output, and more diverse to increase product range. The equipment eventually becomes too large for individuals to house, and too expensive for them to afford individually. The wider product range also becomes too complex for one person to have all the skills needed. The solution is building or acquiring larger and more dedicated production spaces, and have multiple owners with different skills who work together to make what is needed.
There are two approaches to working together: distributed and centralized. In the distributed approach, each person owns one or a few pieces of equipment. A production request is then distributed as needed to the people and their equipment, who do their parts to make the product. In this approach, ways are needed to settle up the resources and costs in meeting each other's needs. In the centralized approach, the equipment is in one or a few larger places, with shared ownership. A mixed approach is also possible, part distributed and part centralized.
As production grows beyond personal needs, the network can start to operate like a commercial workshop, building items for outside customers in addition to the local community. At the hobby and home-use scale, participation is mostly part-time, and the owners' main support is from conventional work. As production grows to commercial levels, some of the people can transition to full-time self-employment and have a large part of their physical needs supplied from within the network.
- The "Makernet" (Section 6)
We define a Location as an area small enough that people can easily travel to work in person, and exchange products with each other. An example would be a US metropolitan area. In the Personal Production example, one location can then contain multiple home and dedicated sites where work is done. Since modern communications are worldwide, the network does not have to be restricted to a single location. So our second example extends the distributed approach to multiple locations in different parts of the world.
Production sites are then network nodes that communicate by means such as the Internet. We call it a "Makernet" because the network collectively makes things. Owner-operators and customers can communicate with each other by any available method, including electronically. Smart production machines can also communicate with each other directly, or be controlled by people and software remotely.
Shipping physical items long distances involves significant cost, energy, and time. Therefore makernet production will tend to be concentrated on items with higher value to mass ratios. At the high end, electronic design files for new machines and products have no mass, and can be sent anywhere at low cost. They are well suited for a worldwide network. At the opposite end crushed stone for construction is heavy and cheap. It will tend to be limited to a single location near the quarry rather than being shipped long distances.
In a well-developed makernet, a customer may submit a product order to the network through a website. Production requests are then routed to various network nodes, who supply the materials, fabricate parts, and do other necessary steps. As each step is completed, payment is routed to the node. Eventually the finished item is delivered to the customer. Production nodes can contain any number of machines and processes. Since new machines can collectively be made by already existing nodes, a single seed factory starter set can in theory extend the network everywhere over time.
For node operators, modern communications allows people to work remotely from the actual equipment. This has potential advantages in commuting cost and flexibility in hours and location. The Personal Production example is a local community at a single location. So much of the interaction can be in person or by voice. The Makernet approach emphasizes software and networking, with people doing things asynchronously. New starter sets would tend to be optimized for size and mass, so they can be shipped farther at reasonable cost.
- Industrial Production (Section 7)
The goal of this example is large-scale production of more specialized products, much like conventional factories. It begins with smaller-scale commercial networks like the previous examples, or starter sets specifically designed to grow to industrial scale. In either case, growth by self-expansion should be a less expensive way to reach the desired production capacity. In one version, the starter sets are mobile, and are used like construction equipment. They are brought to a new industrial site and start making parts and equipment. Once the new factory reaches sufficient capacity, the seed factory components move on to the next project. In another version, the commercial-scale network grows in place to the larger industrial scale.
Individual industrial sites grow towards scale and efficiency rather than diversity of outputs. This is partly because it gets harder to fit all the large equipment and operations in one place. Starter sets and small-scale production favor flexibility in what they can make, at the expense of being less efficient. At industrial scales, the the individual machines can be optimized for their specific tasks. They are then easier to automate because they each do fewer different things.
Industrial-scale sites can still operate as a network, and make things for each other as needed. However, the larger scale output goes well beyond the original owner's needs, or even the general needs of a given location They can then sell products directly to a wider market. They can also sell production shares to people who otherwise don't participate in site operations. Production shares entitle people to a share of the products made, either without further payment, or just to cover inputs like raw materials. An example of this would be a large-scale automated farm, whose shareholders then get a portion of the food produced. Industrial-scale sites can use conventional factory financing via capital markets if internal funding is not sufficient.
- Remote and Difficult Locations (Section 8)
This book's last example is production in remote locations, or where environment conditions are difficult or extreme. In such locations, the benefit of self-expanding systems is not having to bring all the infrastructure with you. Instead, you bring a starter set and produce most of the rest locally. Examples of remote and difficult locations on Earth include deserts, polar regions, and the ocean surface. More extreme locations include high altitudes, ice caps, deep underground, or underwater. Current civilization mostly occupies a thin layer within 30 meters of about 13% of the Earth's surface. Extending it horizontally to the remaining area, and vertically beyond the first few meters would vastly increase available space and resources. The key to this is affordability, which self-expanding systems can help provide.
The more difficult and remote the location, the higher the cost of transporting everything there, and the more incentive there is to use a small starter set and build locally. High levels of automation and productivity can make it possible to live and work in such locations by providing the infrastructure and resources needed. Remote controlled operation allows building up capacity without necessarily living there. For example, the Sahara Desert may be a great place to make solar panels, because of abundant sunlight and sand as inputs. But people may not want to live there full time. Working remotely can allow that choice. Conversely, if technology enables living in a remote location, it might be desirable for some people. Since other people don't live there, land would be cheap. It might be scenic or have other features people would prefer.
- Space (Volume II)
The most remote and difficult places are regions in space beyond the Earth. Synchronous orbits, where many satellites operate, are twice as far as any two points on Earth can be, and the farthest known Solar System orbits extend ten million times farther than that. The space environment has a number of difficult or hostile conditions for people and other living things. This includes lack of breathable atmosphere and water, high radiation levels, and low gravity. On the other hand, space contains large amounts of untapped energy and material resources. If seed factories and self-expanding production can help make these resources accessible, the potential gain for civilization is enormous.
Since the conditions are so different, and involve different engineering specialties and designs, we reserve discussion of their examples to a second volume on Space Systems Engineering. The examples for space are linked to the previous ones for Earth om a couple of ways. First, ndustrial and in some cases remote and difficult locations are needed make and operate transportation to reach space. Second, the first seed factories to use in space must be designed and built down here, and some of their operation can be remotely controlled from Earth.
- Conceptual Design Notes (Section 9)
The final section of this volume is a collection of material that is (a) too new or not organized enough for the main text, or (b) is too detailed and would interrupt the flow of the discussion. The ideas we present in these books are a work in progress, so the final section also serves as a kind of "engineer's notebook" to record that work. In the past, such notebooks were on bound paper or pages in file folders. Larger sketches and drawings were kept in drawers or rolls. Only a few people could access them and collaborate. The modern approach is a shared electronic database where everyone has access to the full contents. Our work is open-source and we encourage collaboration. So the notes in Section 9 form part of our database. Material that is not suited to the Wikimedia format will be stored elsewhere and linked from that section, and from other parts of both volumes.