Introducing JSON-AI

4 min readOct 29, 2018

Situation: Interviewing with Google for a dream job
I was in a ‘whiteboard’ interview for a dream job: a type of ‘professional creative+technology brainstormer’ where I could freely imagine, prototype, and test new AI innovations for Google’s flagship product AdWords. My interviewer asked me to write a schema for a valet company’s new service. I realized in that moment I didn’t have a good understanding of schemas and it obviously felt very important to know for the role I was interviewing for.

What is a schema?
In short, it is a document that contains definitions of things, and anyone that wants to guarantee that the information in their own document has a mutually-agreed-meaning can refer the information in their own document to this third-party definition. This system of creating reference documents has incredible capabilities to product societal benefit, a famous example would be the Dewey Decimal System, a 140-year old system which was the de-facto standard in my childhood:

The Dewey Decimal Classification (DDC), colloquially the Dewey Decimal System, is a proprietary library classification system first published in the United States by Melvil Dewey in 1876.[1] Originally described in a four-page pamphlet, it has been expanded to multiple volumes and revised through 23 major editions, the latest printed in 2011

Before the age of Wikipedia, the user experience of the Dewey Decimal System as a schema of book location was actually considered a marvelous convenience of society, a public library where treasures of knowledge were shared among society. But in order to use a library, I had to:

1. Physically walk or ride my bike to the local library (when I was old enough).
2. Locate the drawer in the correct aisle of a Library Catalog to find the name of the book or the author I am searching for, flip through the hundreds of sticky cards.
3. Upon finding the right card, locating the the dewey decimal id (index) of your book from the card, you would copy it to a piece of paper.

4. Then you have to locate the book in a maze of stacks of books!

It was kind awesome at the time, browsing through books and quite often a new book you hadn’t thought of would grab my attention from the cover art or title.

I didn’t get the job at Google. It was the first question of the interview and once I faced my lack of knowledge in designing schemas I felt ‘off’ the entire rest of the interview.

How are schemas used today?

After not getting the job I became fascinated with machine teaching, schemas, and the experience of listening to coworkers at previous companies of employment who repeated the mantra: “All knowledge management systems suck!”

I thought, how cool would it be to build a schema for knowledge, but that could do better, could provide better user experiences than currently exists? Currently, everyone uses schema.org and creating a variation on that would be a great learning experience. Here’s the description from schema.org’s website:

Schema.org is a collaborative, community activity with a mission to create, maintain, and promote schemas for structured data on the Internet, on web pages, in email messages, and beyond….
Founded by Google, Microsoft, Yahoo and Yandex, Schema.org vocabularies are developed by an open community process, using the public-schemaorg@w3.org mailing list and through GitHub.

The problem with schema.org is that it is an oligarchy of knowledge, a few companies control it, and the taxonomy is centralized. Websites add schema.org metadata to their website’s html and javascript, because it lets Google generate rich knowledge cards in search experiences.

In my job interview I couldn’t demonstrate what I could do given the opportunity to learn at Google, but together with Ken Miyachi and Reagan Karnes, JSON-AI was born:

JSON-AI is your organization’s Dewey Decimal System for knowledge that you get to define using natural language.

We at Intelligence.AI are helping making it just a bit easier to create knowledge about knowledge. Our inspiration is the Microsoft Machine Teaching Research Group, who are doing amazing work!