The ontology formulator immutably records the original expression in natural language on a public blockchain (on-chain). The original data is a string of text in natural language (e.g. JSON file) and is stored in the cloud on a separate centralized server whose exact storage location on the Internet is cryptographically encrypted and identified with an IPFS address. IPFS Protocol ensures[i] that the original data is stored in a distributed network so it’s immutable. The blockchain will only keep basic information about the origin of the data such as authorship, time/date, and its IPFS address (i.e., collectively we call this, provenance).
The corresponding meta data (e.g., canonicals) will then be collected separately and grouped and stored into the AI-specific ontology database running on an off-chain server. For example, the Mind AI engine parses an expression in natural language into their constituent keywords which will then be formulated into a canonical.
This redundant and yet symbiotically bifurcated data storage mechanism serves two functions:
- it ensures immutability of the original data in natural language format so that the canonicals can be reconstructed if the meta data is compromised, lost, damaged, or hacked
- the off-chain treatment of the data processing on a centralized server will ascertain AI systems (e.g., Mind AI) to possess a rapid processing speed, which currently cannot be achieved using a blockchain.
[i] On IPFS, see IPFS documentation. IPFS Docs. July 5, 2021. https://docs.ipfs.io/.