The Principles of Cybernetics
Cybernetics is a fascinating field at the intersection of biology, technology, and information theory that delves into the intricate dance between systems and their environments. It's the science of control and communication, the study of how systems adapt and learn, and the guiding force behind everything from self-driving cars to the human brain.
Determining the single "founder" of cybernetics is a bit nuanced, as the field emerged from the ideas and contributions of several researchers and thinkers throughout the 20th century. However, two important figures often credited with laying the groundwork and solidifying the concept of cybernetics are:
Norbert Wiener: An American mathematician and philosopher, Wiener is widely considered the "father of cybernetics." He published the seminal book "Cybernetics: Or Control and Communication in the Animal and the Machine" in 1948, which outlined the core principles of the field and brought together ideas from various disciplines like mathematics, engineering, neuroscience, and physiology.
Warren McCulloch: An American neuroscientist and philosopher, McCulloch played a crucial role in shaping the theoretical underpinnings of cybernetics. His landmark paper "A Logical Calculus of the Ideas Immanent in Nervous Activity," co-authored with Walter Pitts in 1943, explored the use of Boolean logic to model the human nervous system and laid the foundation for information processing concepts within cybernetics.
While Wiener and McCulloch stand out for their significant contributions, it's important to recognize the collaborative nature of early cybernetics research. Other notable figures who helped develop and shape the field include:
Arturo Rosenblueth: Mexican physiologist and cyberneticist, who collaborated with Wiener on early theories of control and communication in living systems.
Claude Shannon: American mathematician and information theorist, whose work on information theory significantly influenced the understanding of communication and control in cybernetics.
Margaret Mead: American anthropologist and cultural theorist, who explored the potential of cybernetics to understand and analyze human behavior and social systems.
As with many things in science and technology, it turns out that cybernetics was the act of finally naming a body of work that had built up over time, once it had reached a point where it was ready to become its own discipline. Wiener’s book drew on elements of his own research and that of others during World War II to automate shooting down enemy aircraft like new high altitude bombers and V2 rockets to protect London. That work sparked interest in one of the most important concepts of modern times, feedback loops.
Feedback Loops: Nature's Symphony of Control
Imagine a thermostat diligently adjusting your home's temperature. It senses the actual temperature (input), compares it to the desired setting (reference), and adjusts the heating or cooling (output) to achieve equilibrium. This elegant dance is the essence of a feedback loop, the fundamental building block of cybernetics.
Feedback loops come in two flavors: positive and negative. Positive loops amplify changes, like the snowball effect. Negative loops, like the thermostat example, counteract deviations from the desired state, promoting stability. These loops weave a dynamic tapestry, allowing systems to adjust, learn, and thrive in ever-changing environments.
Feedback is found in a few different places. A good modern concept might be sensors in a phone or 3D printer. When a printer moves up on the Z axis, there’s a bolt that keeps it from moving too far. The act of adding that was the result of a feedback loop - the print head slammed into something, or flew off the screw at some point. Generations later, we move from mechanical feedback channels to something like what the Bambu printers have, where there’s AI doing spaghetti detection on a model or the printer detects that it’s out of filament so stops a print. We learned using those feedback loops and evolved the product. Just like ecosystems in nature learn, which then becomes a component of homeostasis.
Homeostasis: Maintaining the Inner Balance
Biological systems, from humble single-celled organisms to complex humans, constantly strive for homeostasis, a state of internal equilibrium. Whether regulating body temperature, blood sugar levels, or hormone balance, our bodies hum with feedback loops, ensuring optimal functioning. Cybernetics borrows this principle, designing systems that self-regulate and adapt to maintain stability within their own boundaries.
Homeostasis wasn’t really a part of the original theories of Cybernetics. William Ross Ashby invented the Homeostat in 1949. He was a British psychiatrist and his remarkable machine, stands as a pioneering example of early artificial intelligence and adaptive systems. The Homeostat wasn't designed to mimic human physiology, but rather to exhibit adaptive behavior through a network of electronic circuits. It consisted of four units:
Fourteen electromechanical elements: These acted as "muscles" for the machine, performing actions based on received signals.
Ten relays: These functioned as "nervous tissue," processing information and sending instructions to the elements.
A photocell: This served as the machine's "eye," detecting changes in light and feeding that information into the system.
A source of random noise: This element introduced an unpredictable factor, mimicking the inherent randomness in environmental conditions.
By interacting with its environment, the Homeostat could learn and adapt its behavior. The photocell and random noise provided inputs, while the relays and elements produced outputs. Through a complex web of feedback loops, the machine could adjust its actions to achieve stability and maintain certain desired conditions within its light-sensitive environment.
Ashby's Homeostat, despite its basic components (which meant he could build it for 50 pounds), demonstrated several groundbreaking principles:
Adaptive behavior: The machine could learn and adjust its actions based on feedback from its environment, a significant step in early AI research.
Emergence: Complex behavior arose from the interaction of simple components, showcasing the potential of bottom-up approaches in understanding complex systems.
Cybernetic principles: Homeostasis, feedback loops, and information processing were central to the machine's operation, highlighting the early application of cybernetic principles in artificial systems.
When Ashby flew to New York to speak at one of the Macy conferences, many of the cyberneticists took umbrage to his machine. At the time, they got lost in the difference between classifying different parts of ecosystems vs the things that lived in those environments, and what learning actually meant. Wiener happened to not be at that specific conference and when he read Ashby’s research, he immediately included it with his own - and homeostasis has been a core concept of cybernetics since.
The Homeostat's impact extended beyond the world of machines. It influenced research in fields like artificial intelligence, systems theory, and even psychology, offering valuable insights into the dynamics of adaptive behavior and control. While subsequent advancements have led to far more sophisticated AI systems, Ashby's pioneering work with the homeostat remains a cornerstone in the history of cybernetics and artificial intelligence.
Information and Entropy: The Dance of Order and Chaos
Cybernetics acknowledges the ever-present tension between information and entropy. Information represents order, structure, and the potential for control. Entropy, on the other hand, is the relentless march towards disorder and decay. Cybernetic systems actively process information, using it to counteract entropy and maintain their ordered state. Think of it as a constant battle against the inevitable decline, where information acts as the shield against the encroaching darkness of chaos.
Understanding the interplay between information and entropy in cybernetics can feel like a philosophical tango, but using 3D printing as a partner for the dance makes it much more tangible. Let's delve into this dynamic duo, starting with information.
Information: In a 3D printing journey, information plays the role of the choreographer. It comes in different forms:
3D Model: This digital blueprint, packed with data points and instructions, dictates the form and structure of a printed object. It's the blueprint for a desired reality. That reality might require the acknowledgement of silly little things like gravity (you know, 'cause it's a law and all). But ultimately the slicer helps take the model to a gcode (or whatever proprietary format) on the printer, which includes some of the settings for the software to interpret.
Software Settings: Parameters like printing temperature, layer height, and infill density fine-tune the printing process, ensuring the information in the model translates accurately into physical form.
Sensor Feedback: Throughout the printing process, sensors gather information about temperature, filament flow, and build plate adhesion. This real-time information allows the system to adjust and maintain control,preventing chaotic failures.
Entropy: Entropy, meanwhile, is the ever-present counterpoint, the force of disorganization seeking to disrupt the information's dance. Think of it as the clumsy guest at the tango, always threatening to trip up the graceful steps:
Filament Degradation: Over time, even unused filament can degrade, losing its structural integrity and affecting print quality. Entropy is slowly chipping away at the potential for creation.
Calibration Drift: Environmental factors like temperature and humidity can misalign the printer's internal mechanisms, introducing errors and imperfections into the print. Entropy is whispering chaos into the gears.
Filament Jams: Blockages and tangles can halt the printing process, disrupting the flow of information and preventing the complete expression of the model's form. Entropy is throwing a wrench into the well-oiled works.
The key to successful 3D printing lies in the constant battle between these two forces. Information, driven by the model, settings, and feedback, strives to maintain order and structure. Entropy, lurking in the materials, environment, and potential malfunctions, seeks to sow disorder and impede the process.
This dynamic struggle plays out in several ways:
Feedback Loops: By constantly monitoring print progress and adjusting settings based on sensor data, the system actively combats entropy's disruptive potential. It's like the tango partners sensing each other's movements and adapting to maintain the flow.
Error Correction: Advanced printers might utilize software algorithms to detect and compensate for errors introduced by entropy. They become masters of improvisation, turning potential stumbles into graceful recoveries.
Maintenance: Regularly calibrating, cleaning, and optimizing the printer keeps entropy at bay, ensuring the information can flow freely and consistently. It's like polishing the dance floor to prevent unexpected snags.
Through this constant interplay, the 3D printer emerges as a fascinating example of a cybernetic system. It harnesses information to fight the pervasive force of entropy and brings a digital blueprint to life in the physical world.
Ultimately, the universe will collapse into chaos. That is the fate of all, as Galactus could have told us had we asked. But on a small scale, we can control that. The best way to think of this is those usage statistics we submit. The Bambu printer keeps getting better because it sends detailed information about jobs that completed successfully or failed back to Bambu. Apple asks for anonymized information to evolve their products. Most vendors now collect information and try to improve products based on what users teach them. We try to restrict failures with unit tests, ever more sensors, etc.
Emergence: The Whole is More than the Sum of its Parts
Complex systems exhibit a fascinating phenomenon called emergence. It's the idea that the properties of a whole system cannot be simply predicted by the sum of its individual parts. Think of an ant colony - individual ants are simple creatures, but their collective behavior exhibits impressive intelligence and adaptability. Emergence arises from the intricate interactions within a system, a testament to the power of the whole that transcends the limitations of its individual components.
Let's see how this plays out in 3D printing:
Individual Components: A 3D printer consists of various parts like the printing head, extruder, build plate, and control software. Each component has its own specific function and interacts with others in a predefined way.
Interacting Parts: When you initiate a 3D print, these individual parts begin to work together in a coordinated manner. The software sends instructions to the printing head, which moves across the build plate, laying down melted filament in precise layers. This interaction between the components and the environment (the filament) is where the magic happens.
Emergence of a 3D Object: As the layers accumulate, the printer "emerges" a completely new object that wasn't present before. The object's form, functionality, and even aesthetic qualities couldn't have been predicted by simply looking at the individual parts of the printer. It's the result of the complex interplay between the printer's components, the filament, and the software's instructions.
This emergence of a functional object showcases several key aspects of cybernetics:
Feedback Loops: The printer constantly monitors its progress through sensors and adjusts its actions based on the feedback (e.g., filament feed rate, bed temperature). This dynamic feedback loop allows the system to adapt and achieve the desired outcome.
Information Processing: The software plays a crucial role in processing the 3D model data and translating it into instructions for the printer's movements. This information processing is at the heart of cybernetic systems.
Self-Organization: While guided by the software, the printer exhibits a degree of self-organization in optimizing its movements and material deposition to achieve the best results. This ability to self-organize is another key trait of cybernetic systems.
The 3D printer is just one example of how emergence manifests in cybernetics. We can see it in other systems like self-driving cars, robots that learn to navigate complex environments, and even in biological systems like the human brain.
Adaptability and Learning: Embracing Change
Cybernetic systems are not static; they are designed to adapt and learn. By processing feedback and interacting with their environment, these systems can adjust their behavior and improve their performance over time. Just as a child learns from experience, cybernetic systems can refine their responses and strategies, becoming more adept at navigating the complexities of their world.
These are just a few of the core principles that guide studies in a cybernetic fashion. By understanding these foundational concepts, we gain a deeper appreciation for the intricate dance between systems and their environments, the elegance of feedback loops, and the relentless pursuit of order amidst the ever-present threat of chaos. There’s a maturity model. That printer with the mechanical stops or sensors to know when to start extruding plastic are one part of that. Practically every 3D printer on the market does that (otherwise the print head would really jsut be a 3D pen). But there’s another phase of maturity after that.
The Flashforge was the first 3D printer I saw that had auto-leveling, a built-in camera, and sensors in new places that went beyond the reprap standard. Turns out it didn’t really auto-level and the camera was to watch the print jobs, not feed ML models. You levelled one point and it did 8 more (later 30 total). More sensors provided more potential feedback loops, automation (and thus productivity), and kept failures at a minimum for the technology of the time. I actually wrote an article about where I thought we were at with 3d printing right around that time here: https://www.bootstrappers.mn/post/it-s-time-for-3d-printing-to-evolve-from-the-hobbyist-market.
Then comes the Bambu. Now they’re puting tinyML type models to process data from the camera, adding more sensors, like to detect when the printer runs out of filament, and learning from the shortcomings of former products. These are innovation cycles based on feedback systems. From 3D printers to robots to ecosystems, cybernetics provides a lens through which we can understand the complex interplay between systems and their environments. Like making the printers enclosed so parts don’t curl up due to the outside air.
Cybernetic principles not only guide the development of sophisticated technologies but also offer valuable insights into the workings of our own biological systems and the intricate dance of life itself. And yet, when I took product management classes, they simplified the entire study of systems into a very basic “send surveys” or “visit customers” approach. Perhaps this is why so many entire categories of products never seem to evolve - or at least they don’t until they’re forced to. For example, take password managers. They’re pretty much all encrypted sqlite3 databases. Some have added other types of secrets out of necessity, like passkey support. Some have also built in other options, like SSH agents, because their feedback was the buyer. But this just helps them appeal to people in IT. Others just copy each other, and so appeal to a subset of the IT crowd.
Going beyond the basic surveys means a deeper thinking about the problem space and then coming up with a number of hypothesis about what is next. Survey users, and analyze the concepts, but ultimately there’s also a modicum of trusting your gut that you’re pushing the envelope. That’s when we filed our patent. And as we surveyed other tools and the features they have, we really started to find that most had places where they did cool things. Maybe even things we didn’t want to do, despite thinking they were cool. Heck, we even became paying customers to use some of those cool things. But our primary focus is to change the mentality of the entire industry. We want every vendor (and there honestly aren’t that many) to think about secret storage in a new and innovative way. Because threat actors certainly are.
<rant>Our goal is to care about humans and up the ante to help keep them secure. The best UX is a sense of security, like a parent wrapping their arms around a child to keep them safe. We hope you'll join us for a better future, where we're all hooked up with sweet productivity gains that protect our privacy while implementing learning systems so we can just keep innovating (until the universe implodes, but that's hopefully the only thing that can make us stop - and if we keep at the incremental productivity gains as a species, maybe we'll be able to jump to another universe before it happens). </rant>