The most famous line in Jurassic Park is biologically backwards. Tyrannosaurus Rex had a stereoscopic visual field wider than a hawk's, visual acuity roughly 13 times sharper than a human's, and depth perception comparable to a military rangefinder. Standing still would not have helped. The science of how we know, drawn from the geometry of the skull itself.
The Archive
A model trained to detect skin cancer learned to look for rulers instead. The explainability problem is not a glitch. It is Goodhart's Law running at machine speed — and the gap between what AI systems can do and what we understand about how they do it is where the most consequential work in the field is now happening.
In 2012, a neural network cracked image recognition in a way that reshaped the entire AI field. Andreas Møgelmose traces what neural networks actually are, how convolutional architectures build understanding in layers, and why a cancer-detection model that learned to spot rulers instead tells us something important about where the field still cannot go.
Fusion's greatest achievement is not the plasma temperature or the confinement geometry. It is seventy years of open, international collaboration across the Cold War divide — and what that model reveals about how civilisation-scale problems actually get solved.
On the air
Navigate by field
Cross-cutting principles
A process in which the output of a system is fed back as input, either amplifying or suppressing the original signal. Feedback loops govern stability and instability across physics, biology, climate, and economics.
The way systems behave differently depending on their size or resolution. Scaling laws reveal which properties grow proportionally, which change radically, and why the rules that govern one scale often break down at another.
The quantifiable property of a signal or message that can be encoded, transmitted, and decoded. Shannon entropy provides the mathematical foundation. DNA, neural signals, and digital communication are all physical implementations of information processing.
The process by which heritable variation, differential reproduction, and natural selection drive cumulative change over time. The principle applies most precisely in biology but appears in analogous forms wherever variation and selection act on replicating systems.
The process of finding the best solution within a defined set of constraints. Nature and physics are full of optimisation: Fermat's principle of least time, the principle of least action, natural selection, and algorithmic search are all instances of the same underlying logic.
The fundamental limits on what can be known or predicted. Heisenberg's uncertainty principle makes this a property of physical reality, not merely of measurement. Statistical mechanics and information theory describe how to reason carefully within those limits.
Applied domains
The field concerned with building systems that exhibit intelligent behaviour.
The application of biological knowledge to engineering.
The science and engineering of the universe beyond Earth.
The capacity to do work, and the central organising challenge of modern civilisation.
Systems defined by their connections rather than their components.
The science and practice of making things at scale.