Skip to content

AI DSL requirements

Jump to bottom
kabir edited this page Feb 24, 2021 · 14 revisions

AI-DSL ontology defines data and task types to be used by AI-DSL. Requirements for the ontology are shaped by the spope and specification of the AI-DSL itself. Here is the short summary of such general AI-DSL requirements1:

Place in SNet ecosystem discuss

  • Each SNet Layer 2 system will be represented within the AI-DSL via a system-specific sybsystems to match and coordinate agents;
  • AI-DSL will be used for explicit support of the notion of AI agents systems launched for coordinated interoperation in terms of:
    • 'service fleet management', requiring interface for manually or automatically managing many interconnected SNET AI agents;
    • 'decentralized service mesh', supporting services calling other services in the platform in a decentralized manner;
  • SNet AI-DSL wil integrate into OpenCog Hyperon (and possibly other SNet Layer 3 technologies) and used to guide internal representation of algorithms in a network populated by heterogenous inter-operating AI agents;

High level functionality discuss

  • An AI service ontology for easier service discovery;
  • A full dependent-type language for AIs to describe capabilities to each other
  • Deep integration of the AI-DSL into the Cardano Plutus smart contract framework

Roadmap

Long term

  • 2021:
    • Alpha version of AI-DSL for Cardano-based Singularity NET agents;
    • Service ontology (as part of AI-DSL ontology) for easier discovery or AI agents based on various criteria;
  • 2022:
    • integration of AI-DSL into routine operations of Cardano portions of the network;
    • extension of SL2 systems using AI-DSL;
    • service quality evaluators, interfacing with the platform via AI-DSL;
    • matchmaking bots leveraging AI-DSL to automatically match requests with agents capable to fulfill them;
  • 2023-25:
    • AI tool for converting natural language descriptions of agents to formal descriptions of agents on AI-DSL

Short-term for year 2021 discuss

First phase

  • Importation of existing data-type ontologies into AI-DSL framework, covering common data types such as images, text, videos plus some forms of relational data;
  • Initial formalization of compositional laws for key aspects of financial, temporal and computational costs
  • Initial formalization of compositional laws for simpler aspects of quality of results
  • Test suite for the partial AI-DSL implemented (see section)

estimated time frame -- about 6 months from the start

Second phase

  • Importation of relevant domain-specific ontologies into AI-DSL framework, e.g. low hanging fruit will be biomedical and finance domains where existing ontologies are fairly mature
  • Complete the formalization of compositional laws for financial, temporal and computational costs
  • Complete the formalization of compositional laws for quality of results (this will be a significant fraction of the work in this phase)
  • Extended test suite for the AI-DSL
  • Initial version of end-user language for AI-DSL
  • Integration into SingularityNET-on-Cardano platform (assuming that by the time Phase Two comes around, the porting of the SingularityNET marketplace to Cardano will have been completed)
  • Description of a set of 5-10 useful SingularityNET agents in the AI-DSL as reference cases

estimated time frame -- about 6 months after finishing the first phase

Footnotes

[1] : initial sources of general AI-DSL requirements:

Clone this wiki locally