About

mldata.opendata.ai: The Provenance Sandbox

Supported by Link Digital | Powered by CKAN | Incubating the Objective Observer Initiative (OOI)

Welcome to mldata.opendata.ai. This portal is an experimental data infrastructure sandbox developed by Link Digital. It has been specifically established to support the MLCommons Datasets Working Group by providing a live testing ground for next-generation data provenance, metadata standards, and defenses against AI data poisoning.

While currently operating as a standard, highly-optimized CKAN catalog, this portal is the staging environment for a much deeper technological shift: the integration of Chronovertical Data via the Objective Observer Initiative (OOI).

The Crisis in Machine Learning: "Flat Data" and Loose Entropy

As Generative AI scales, the industry's primary bottleneck has shifted from compute power to data integrity.

Currently, ML datasets rely on "flat data." Because traditional digital networks treat information as "weightless," metadata, timestamps, and origin signatures are entirely software-based. This means they can be easily spoofed, stripped, or hallucinated. As the internet fills with AI-generated synthetic data, training models on these contaminated datasets leads to a mathematical degradation known as Model Collapse.

To secure the future of AI, we cannot rely on hackable software ledgers. We need absolute, mathematically undeniable "Ground Truth."

Baseline Testing: The Source Cooperative Harvest

To simulate the rigorous demands of high-fidelity, large-scale "stored data" use cases, you will notice we are actively harvesting complex geospatial, climate, and demographic datasets from Source Cooperative.

Why start here? Cloud-native Earth observation data provides the perfect heavy-duty baseline. It allows us to stress-test how robust storage architectures handle massive data structures and standard metadata ingestion (such as the Croissant format). Establishing this reliable "Stored Data" baseline is a prerequisite before we can introduce the rigorous hardware requirements of Constructed State Preservation.

The Horizon: The Objective Observer Initiative (OOI)

This is where the paradigm shifts from standard data management to information physics.

Link Digital is currently incubating a provisional patent for the Objective Observer Initiative (OOI)—a 100% solid-state, hardware-enforced causal computing architecture. The ultimate goal of this portal is to bridge standard ML training pipelines directly into the OOI physics engine.

Through OOI, we are introducing Chronovertical Data to the ML ecosystem.

  • What is it? Chronovertical data is not secured by a standard cryptographic hash. It is data that has been physically and mathematically phase-locked into a solid-state memory matrix at the exact millisecond of its creation.
  • How does it work? Using a physical "escapement mechanism," data is mathematically confined and subjected to thermodynamic work. It is only stored when perfect triadic network alignment is achieved, creating a zero-knowledge proof of its exact causal history.
  • Why does it matter for ML? When an ML model ingests Chronovertical Data, it ingests information with zero loose entropy. Adversarial data poisoning becomes mathematically impossible, because an injected hallucination cannot generate the correct thermodynamic friction required to pass through the hardware tunnel.

Current Status & Sandbox Roadmap

We believe in radical transparency as we build.

  • Phase 1 (Current): The portal is a standard CKAN instance. We are actively testing the harvesting, display, and API distribution of Source Cooperative data to ensure frictionless integration with standard ML workflows. (The physical OOI backend is not currently connected).
  • Phase 2 (Upcoming): We will begin bridging this CKAN instance with the OOI proof-of-concept Python scripts. This will allow researchers to observe the simulated "Thermodynamic Hash" and chronovertical metadata alongside standard dataset files, testing the low-entropy compute, network, and storage claims.
  • Phase 3 (Hardware Realization): Delivering true zero-knowledge, zero-entropy chronovertical data directly to ML training pipelines via the OOI solid-state causal kernel.

Get Involved

We recognize that bridging theoretical physics, solid-state hardware, and machine learning pipelines is a massive leap. We invite the MLCommons community and AI researchers to explore the sandbox, utilize the CKAN APIs, and collaborate with us. We are establishing the plumbing today so that tomorrow, we can deliver datasets guaranteed by the laws of causal physics.

To learn more about Link Digital's commitment to open data and ML infrastructure, visit linkdigital.com.au.