Artificial Intelligence is learning everything — except what things actually mean. In the global race to feed machines with more data, we’ve quietly lost the most essential part of science itself: measurement.
Behind every elegant neural network, every predictive model, lies a mountain of numbers — and most of those numbers no longer come with units, uncertainty, or context.
A number without a unit isn’t knowledge; it’s noise.
When machines train on this noise, they start believing that 5 = 5, whether it’s 5 kilograms, 5 kilometers, or 5 seconds. That’s not intelligence — that’s ignorance at scale.
The Hidden Flaw in the Data Revolution
The problem began innocently enough.
To make data easier to share, scientists, companies, and governments stripped away the messy parts — some of the “metadata” — that explained how the data was measured, in what units, under what conditions, and with what precision.
In a world of trillions of data points, those little details seemed expendable. But in doing so, we’ve created an AI ecosystem that’s dimensionless or contextless.
When Measurement Mistakes Go Galactic
In 1999, NASA lost the Mars Climate Orbiter — a $327 million spacecraft — because one engineering team used pound-seconds while another used newton seconds. The units didn’t match; the spacecraft disintegrated in the Martian atmosphere.
AI won’t crash a spacecraft — it might crash our confidence in science itself.
The Generation of Erroneous Learning
Modern AI is trained on oceans of data from sensors, social feeds, satellites, and labs. But without context, these systems start internalizing wrong patterns — numerically consistent but physically meaningless.
Researchers at Google call this the “Data Cascade” effect: small data missteps that snowball into huge model failures downstream. Harvard scientists have a name for it too: “underspecification” — when multiple models fit the data but make wildly different predictions in the real world.
The result? An AI generation that looks smart, sounds confident, and learns — all the wrong things.
The Forgotten Science: Metrology
Metrology — the science of measurement — is humanity’s oldest quality control system. It gave us the meter, the kilogram, the second, and even the volt. Every physical constant is a promise: measurements made anywhere in the world will mean the same thing.
AI has broken that promise.
It’s time to put metrology back into machine learning.
What is the Solution? – Introducing the “Semantic Measurement Layer”
A global alliance of data scientists, engineers, and metrologists is calling for a new standard — a Semantic Measurement Layer (SML) — that ensures every number in a dataset carries its unit, uncertainty, and context.
Think of it as a truth-checker for data.
Using existing frameworks like QUDT (Quantities, Units, Dimensions, and Data Types) and W3C PROV-O for provenance, this layer ensures AI doesn’t just learn patterns — it learns physics, meaning, and truth.
Projects like OMEGA-X CSDM in Europe are already leading the way, embedding semantic interoperability into energy data spaces.
Why It Matters
• Climate AI could stop making temperature predictions that mix °C and Kelvin. • Healthcare AI could distinguish mg/kg from mg, saving lives. • Energy AI could stop confusing power (kW) with energy (kWh). • Finance AI could trace data sources and calibration to prevent algorithmic fraud.
Without measurement, even the most ethical AI is still unreliable.
The Way Forward
Step 1: Attach units, uncertainty, and provenance to every numeric dataset.
Step 2: Use open ontologies like QUDT and ISO/IEC 80000 for unit consistency.
Step 3: Adopt documentation practices — Datasheets for Datasets and Model Cards — for context.
Step 4: Regulate contextual integrity under frameworks like the EU AI Act and NIST AI RMF.
This is not red tape. It’s the backbone of trustworthy intelligence.
Restoring Measurement, Meaning, and Truth
We don’t need smarter machines. We need better-measured data.
AI’s Uncertainty Principle is not about quantum physics — it’s about human negligence. We taught machines to think without teaching them what things mean.
Restoring measurement is not a technical upgrade. It’s a moral one. Because without units, uncertainty, and provenance — there is no truth, only computation.
The author is Sumit D Chowdhury, MD Gaia, GreenEarthX & Digital Energy Grid, Independent Director – Metrology Data Private Limited
Disclaimer: The views expressed are solely of the author and ETCIO does not necessarily subscribe to it. ETCIO shall not be responsible for any damage caused to any person/organization directly or indirectly.
