Ontology-First (or "how to speak platform businesses into existence")
...the secret is not in the sauce source, it’s in the recipe pattern
An elemental capability of computers is its ability to count and discern patterns. This concept of “fit-for-purpose” goes beyond computer technology and is present in other areas where objects appear to have an express purpose within their ecosystems.
Notable “technohumanist” Quentin Hardy once observed that every entity within an ecosystem seems to optimize its function: "a chainsaw wants a tree, and computers want to count and identify patterns." At first reading this statement seems odd, but it is simply abstraction being used to identify and apply a single pattern to diverse objects. Humans have taken this a step further by creating complex language, which has expanded possibilities for social organization over time and space. Encoded learning is a prime example of this phenomenon. According to Hardy we continually create and improve ecosystems by discussing and modelling abstractions.
Digital technology has accelerated this ability, allowing us to touch many things at once, leading to more opportunities for innovation and collaboration. The emergence of APIs, or Application Programming Interfaces, is a good example of such optimization in action. In addition to enabling seamless system integration, APIs allow for the easy sharing of data and functionality among internal and external developers, opening up new avenues for collaboration. This enables developers to continually build upon each other's work, resulting in a portfolio of increasingly diverse digital services that can reach an ever-expanding audience.
The progression of abstraction seems to be driven by an invisible force unleashed by time, which we as humans struggle to keep up with. Comprehension and especially appreciation always come “after the fact.” Long after the discovery of the pattern, as it takes time for appropriate terminology to develop that can facilitate communication of new patterns and ideas. For example, the “pattern” for APIs was first proposed in 1968, which the inventor, Douglas Mcllroy, called “pipes.” It only became known as APIs in 1991 with the release of the first version of the Windows API (Win32); subsequently giving rise to further abstraction and patterns like REST, Cobra and modern web APIs.
The language that is used initially is informal and known only to those close to the discovery or development of the idea (e.g., “pipes” for APIs). As more material is produced and published, new terminology is introduced to the wider world, and with this, an "academic" language develops. This specialized language encapsulates the new knowledge, enabling us to explain and disseminate understanding of these patterns and paradigms. This language is referred to as an "ontology."
Ontologies serve as formalizations of knowledge gained through abstraction and modelling. An ontology is a description (like a formal specification of a program), of the concepts and relationships that can formally exist for an agent or a collection of agents. Ontologies are often equated with taxonomic hierarchies of classes, class definitions, and the subsumption relation.
Modern ontologies are composed of digital avatars (“objects'') that capture the collective learning gained through abstraction. In other words, representing the knowledge as digital objects that have properties, attributes and linkages to other objects. Digital objects have specific traits, such as editability, interactivity, openness, and distributedness, which give them a distinct functional utility. It makes the digital object different from its physical counterpart - the actual thing it represents, which is usually obvious and inevitable (in other words, the physical object is fixed, it is what it is; the digital representation of that object is not).
Interactivity, in particular, enables us to explore a digital object's functions and information without necessarily changing it (interactivity is different from editability, which implies changing the object). Interactivity gives us the ability to make choices and explore possibilities related to a digital object, an ability that is limited and even impossible with physical objects (or with paper-based records of such objects). Take a white (physical) coffee mug that is 12 cm tall, and 8 cm in diameter… it is what it is.
Now take the digital version of that cup… we can shrink its diameter, make it taller, change the color, modify the shape… that is editability. We can do the same for interactivity by exploring (potential) linkages to other objects or data sets: Who can use it? What can it be used for? What happens if you use it to contain tea, or molten lead, or poison? You get the idea.
Although physical objects have a degree of malleability to adapt to changing circumstances, the range of possibilities in the case of digital objects is on an entirely different level. Its key quality is contingent exploration made possible by the responsive and unbundled nature of the digital object. Interactivity unlocks the interior of the digital object, providing opportunities for users to explore alternative courses of action, configurations, and innovations, similar to a user using software to simulate and explore a 3d rendering of his dream house.
This transformative power of digital objects fundamentally changes how we interact with the world. Through visualization and modeling, we can examine objects and their relationships in unprecedented detail, leading to new insights and discoveries.
Such exploration through digital interaction uncovers not just what is, but what could be - extending our understanding of an object's properties, its metadata range, and linkage possibilities. Such explorations enriches the ontology's body of knowledge, making it progressively more comprehensive and precise. Like a scientific theory refined through experimentation, the ontology evolves to capture every verified truth about its domain, becoming an increasingly accurate representation of reality and possibility.
Modern ontologies have evolved beyond simple knowledge repositories into structured frameworks that can be formally encoded and automated. Using languages like the Web Ontology Language (OWL), ontologies can precisely define object characteristics and its interrelationships. These definitions include:
Property characteristics (transitivity, symmetry, etc.)
Relationship cardinalities (one-to-many, etc.)
Logical conditions and constraints
Inheritance hierarchies
Domain-specific rules and axioms
This formal structure enables querying the ontology through languages like SPARQL, allowing precise information retrieval and knowledge discovery. The rigorous nature of these definitions provides the consistency and predictability needed for automation - from code generation to automated reasoning systems.
The ontological ecosystem isn't static - it grows through published vocabularies like WAND and ontology frameworks like MONJO and DEON that provide ready-made building blocks of domain knowledge. Think of these as shared recipe books that everyone can use and adapt. They represent a collective effort to standardize domain knowledge giving us standard ways to describe and understand specific domains.
Through this combination of formal structure, standardized vocabularies, and interactive exploration, ontologies become not just repositories of knowledge, but living systems that grow and evolve with our understanding. They represent perhaps the ultimate expression of the pattern recognition that Hardy identified as fundamental to computing - enabling us to encode them in ways that machines can understand and humans can build upon.
Being formal representations of knowledge and concepts, ontologies offer us something remarkable - the ability to literally 'speak' software applications into existence. By putting "AI on rails" through an ontology-guided language model, businesses can articulate their requirements in natural language and have them automatically rendered as working software applications.
Brilliant perspective on this topic. The way you've connected the dots makes perfect sense.