https://d2ng68cf36kxsk.cloudfront.net/store/public/post/812/post_images/Slide243.JPG">

After going through the philosophical precursors, from Aristotle (but not the Indian philosophers), to Kant, Whitehead, and Peirce, John Sowa arrives at Cyc, the largest and most comprehensive of the digital metaontologies. This was 1995. In 2000, Cyc was still a centerpiece of Sowa’s exhaustive book, Knowledge Representation: Logical, Philosophical and Computational Foundations.²¹²

But by 2006, Sowa was beginning to despair at the “knowledge soup,” where “the best-laid plans” of ontologies, based as they are on people’s “natural desire to organize, classify, label, and define the things, events, and patterns of their daily lives … are overwhelmed by … inevitable change, growth, innovation, progress, evolution, diversity, and entropy.”²¹³

Doug Lenat, then a professor at Stanford University, set out on the Cyc project in 1984. Later, he left academe to found Cycorp. In the words of the Cycorp website, Cyc is “a formalized representation of a vast quantity of fundamental human knowledge: facts, rules of thumb, and heuristics for reasoning about the objects and events of everyday life.”

Today, Cyc “contains over five hundred thousand terms, including about seventeen thousand types of relations, and about seven million assertions relating these terms.” Its top-level architecture consists of individuals (our instances), collections (our concepts), and functions (our relations). Then there is a myriad of “microtheories” covering all manner of domains (weather, time, geography, and such like), marked up in its own Knowledge Representation Language.²¹⁴

Ramanathan Guha worked at Cycorp for a while, and later went to work for Google. Cyc knows stuff like, “A bat has wings. Because it has wings, a bat can fly. Because a bat can fly, it can travel from place to place.” Says Guha, “It’s all the things that a five-year-old knows … Computers still don’t know that.” Computers need to be told common sense, argues Lenat. “Common sense isn’t written down. It’s in our heads.”²¹⁵

Actually, we would respond, there is too much to be meant for it all to be in any of our heads, which is why we write it down, and record via other multimodal means, including computer-recorded text, sound, image, visualization, and dataset. Our records need to be as complex and variegated as the meanings in the structures of materiality to which the records refer – the meanings of nature and history. Computers need to be told these things before we can have them look things up for us. To pass the Turing test (where a user can’t tell the difference between the response of a person and a computer hidden behind a screen),^§MS1.3.2c Lenat correctly says that a machine would have to be told a whole lot of stuff.²¹⁶

Other attempts have been made to create top-level ontologies,²¹⁷ but Cyc has been the most comprehensive. Little wonder that among the corporation’s clients is the US National Security Agency, according to documents leaked by whistleblower Edward Snowdon.

Google has now created its own metaontology in the form of Knowledge Graph. Launched in 2012, Knowledge Graph presents summary text and images about a search topic in a panel on the right side of the screen. A search on “Taj Mahal musician” works because Knowledge Graph knows the difference between the person, the tomb in Agra, and Donald Trump’s failed casino in Atlantic City.²¹⁸ This can only happen when digital text is driven by ontologies. One-third of Google searches now return info boxes from Knowledge Graph. This has reportedly led to a decrease in visits to Wikipedia.

Metaontologies operate at two meta-levels. One level is the distinction (in the terms we have developed in this grammar) between instances, concepts, relations, properties, and such like. At a second level this is filled out with a lot of domain-specific concepts, relations, and properties, such as that people can be musicians, tombs can be in places, and the owners of failed businesses can become presidents. Spatially arranged lists and lists of lists become the ramshackle stuff of digital ontologies. This time the “meta” part is the coverage of everything, including “other,” in anticipation of new discoveries and insights.

The latest project to create a metaontology, a structure for the structures of digitized meaning, is Schema.org, initiated by Microsoft, Google, and Yahoo in 2011. Guha was one of its founders. In the Schema.org metaontology, the world consists of things, which can have names, descriptions, URLs, and images. Then these things can be arranged into item types, including persons, places, events, and creative works.²¹⁹ Over ten million websites now use ontologies marked up in the Schema.org metaontology.

Despite Sowa’s pessimism about knowledge soup, much has come to pass since he and Gruber began the discussion of ontologies in 1995. These are novel forms of meaning in a number of respects. They are multimodal, where with equal ease meanings can be specified that are represented in text, image, sound, or data. However, in tagging structures and their spatial arrangement they are essentially textual, and in a largely unspeakable form where text deviates ever further from speech.

They also operate in a translingual semantics, where translations into different languages do not change the fundamental structures of meaning, many of which are in image or data, anyway. Ontologies and metaontologies are the Esperanto of the digital era.²²⁰

And this now-universal, federated structure of our meanings is governed by stateless, non-profit entities because the state is withering away, and no global state has replaced it. Corporations need shared protocols for meaning, but they can’t do it for themselves. Governance is required, and with the failure of the state, corporations take the lead. Materially as well as ideally, such are the stateless dynamics of the total globalization of meanings in the digital age.

---

²¹¹ John F. Sowa, 1995. “Top-Level Ontological Categories,” International Journal of Human-Computer Studies 43(5–6):669–85.

²¹² John F. Sowa, 2000, Knowledge Representation: Logical, Philosophical and Computational Foundations, Pacific Grove, CA: Brooks Cole.

²¹³ John F. Sowa, 2006, “The Challenge of Knowledge Soup,” pp. 55–90 in Research Trends in Science, Technology and Mathematics Education, edited by J. Ramadas and S. Chunawala, Goa, India: Homi Bhabha Centre.

²¹⁴ www.cyc.com/kb/

²¹⁵ Cade Metz, 2016, “One Genius’ Lonely Crusade to Teach a Computer Common Sense,” Wired, 24 March; Ramanathan V. Guha and Vineet Gupta, 2016, “Communicating Semantics: Reference by Description,” arXiv:1511.06341v3 [cs.CL].

²¹⁶ Douglas B. Lenat, 2016, “WWTS(What Would Turing Say?),” AI Magazine 37(1):97–101.

²¹⁷ Liam Magee, 2011, “Upper-Level Ontologies,” pp. 235–87 in Towards a Semantic Web: Connecting Knowledge in Academic Research, edited by B. Cope, M. Kalantzis, and L. Magee, Cambridge, UK: Elsevier.

²¹⁸ Cade Metz, 2016, “One Genius’ Lonely Crusade to Teach a Computer Common Sense,” Wired, 24 March.

²¹⁹ http://schema.org/docs/full.html#datatype_tree

²²⁰ Bill Cope, Mary Kalantzis, and Liam Magee, 2011, Towards a Semantic Web: Connecting Knowledge in Academic Research, Cambridge, UK: Elsevier.

---

Cope, Bill, and Mary Kalantzis. 2020. Making Sense: Reference, Agency and Structure in a Grammar of Multimodal Meaning. Cambridge UK: Cambridge University Press, pp. 323–326.

Ramanathan Guha’s Schema.org §MS3.4a