←

Article

Orbit

Why every measurement needs a margin of error, and AI rarely gives one

November 15, 2026 · Freethinkers International · Philosophy & critical thinking · 1 min read

[ < 1 minute to read]
Image: Andrei Stiru, Scientias.nl (Dutch)

A Dutch explainer in Scientias.nl visits VSL, the national metrology institute in Delft, where every measurement is traced back to natural constants and always comes with an uncertainty. AI models break that habit: they give a single confident number, identical data can produce different models, and a tiny change in input can shift the answer a lot. VSL is testing how to measure uncertainty from the model, the training data and the new input.

Data scientist Marcel van Dijk warns that no standard method exists yet, and that even the EU AI Act ignores measurement uncertainty. His advice is simple: ask what a model was trained on, and remember that people trust models too readily. Healthy doubt about a number is the start of critical thinking.

Source: Andrei Stiru, Scientias.nl (Dutch) (2026-11-14)
https://scientias.nl/mensen-vertrouwen-een-ai-model-blindelings-waarom-elke-meting-een-foutmarge-nodig-heeft/

Summarised by the FTI newsdesk.

0 0 votes
Article Rating
Subscribe
Notify of
0 Comments
Newest
Oldest Most Voted
0
Would love your thoughts, please comment.x
()
x
×