In Plato’s Republic, Socrates explains that we all resemble prisoners chained inside a grotto, and we are yet to grasp that there is more to reality than the shadows of unchanging ideas projected on the walls of a cave.

Convinced that what we perceive “reflects” reality has enabled humanity to make tremendous progress. Pythagoras unmasked the connection between the inner harmonies of nature and mathematics.

Thales of Miletus freed phenomena from godly intervention. Aristotle introduced scientific analytics. Monotheistic faith shifted the balance towards a mixture of faith and reason, but this was not without institutions like the Inquisition intervening to interrogate Galileo.

Aided by the printing press, the Protestant Reformation declared that individuals were capable of, and responsible for, defining the divine themselves. The birth of the Book meant that knowledge could proliferate faster than the creeds for censorship.

Centralized knowing controlled by the Catholic Church, or the Habsburg-led Holy Roma Empire could no longer edit disavowed ideas. Freethinkers found refuge in societies with advanced publishing industries. Doctrinal ferment followed. Diversity, argumentation, and fragmentation flourished. Discord blossomed. During this renaissance, a new humanism cultivated the “humanities”. Leonardo da Vinci, Michelangelo, and Raphael were admired.

Daring shifts in knowledge occurred with Copernicus’ heliocentric system and Newton’s Laws of Motion. Later, van Leeuwenhoek’s catalogue of the microscopic world unveiled invisible layers of reality. Kant’s “Critique of Pure Reason” proposed a realm of knowing independent of filtration through human concepts. These empires of knowledge led to many great advances including better optical lenses.

Churruca established an observatory at Laventille. On 2 January 1793, he fixed, for the first time, an accurate meridian in the New World. Observatory Street in Laventille, Trinidad, was named for the Churruca Observatory. Einstein would shortly pioneer quantum physics, and Heisenberg and Bohr challenged the long-standing assumptions about the nature of knowledge. Using this knowledge, we have built smart cities, autonomous vehicles, and digital currencies. Next is the Age of AI.

But does AI learn using Plato? The answer is no. Rather, as Wittgenstein would argue, AI learns a language. How do humans learn to use the word “chair”? Not by considering the idea of a chair. Humans learn that it is an object with four legs. But soon we encounter the milker’s chair with one leg, and a Colani chair that has no legs—yet it is a chair that we can sit on. Conversely, a Barcelona daybed is not a chair, even if we can sit on it.  Thus, it is a family of features and their function that make a chair recognizable. Wittgenstein outmanoeuvres Plato in AI.

When AI models learn to distinguish a sheep from a dog, does it start from the idea of a sheep and then recognize all its manifestations? Wittgenstein sets aside the notion of a single essence of things identifiable by reason. This was the goal since Plato. Instead, Wittgenstein advised that knowledge was to be found in the generalizations around similarities across phenomena or “family resemblances”.

His idea was to build a matrix of similarities with overlaps and criss-crossings involving overall similarities or similarities of detail.  He argued that a catalogue of all things with sharp boundaries was mistaken. Wittgenstein contended that the task was to define similar things, even if they had blurred or indistinct edges. It is this idea that has informed AI theories. Not taxonomies.

Wittgenstein wrote: “The meaning of a word is its use in the language”. Wittgenstein recognised that language was not a system of reference. He saw language not as a unitary and comprehensive representation of the world, but as an endless sequence of game-like activities with no unifying essence.

He defined these activities as “language games”. Flying a flag at half-mast is a way of doing language. The act is not a statement of facts. He rejects his earlier picture theory of meaning that holds that language is a mirror of a world made up of discrete facts that correlate to the real world. Wittgenstein recognised that language was a collection of activities in the world. And that these activities can be recognized as language by humans despite their differences.

He uses the analogy of games because games are highly variable and include video games, board games, and sports. While one game may have nothing in common with another, we still identify both as games. Humans see a family resemblance among games; there is no one common feature to all games. Some games share no commonality, but they are connected by a pool of attributes – not by one attribute – that make up the sum of games. The overlapping characteristics are just like family resemblances.

You may not have your aunt’s brown eyes like your brother does, but you may have her black hair like he does. These unorthodox ideas posited that AI’s potential lay partly in its ability to scan data lakes, to learn patterns and features of images in different settings, and to make sense of reality by identifying networks of similarities and likenesses, with what the AI already knew. AI accesses and organises reality differently from humans, and the outputs will alter our perception, cognition, and interaction with the world.