Karl Friston

Born 12 July 1959, York, England

British neuroscientist who developed the free energy principle and active inference, a theory of how living systems maintain themselves by minimising surprise. He also created the standard methods for analysing brain imaging data.

Life

Friston studied natural sciences at Cambridge and trained in medicine in London and psychiatry in Oxford. He spent two years at the Neurosciences Institute in California working with Gerald Edelman, and then moved to University College London, where he became scientific director of the Wellcome Trust Centre for Neuroimaging. He is one of the most highly cited living scientists.

Influences

Hermann von Helmholtz’s nineteenth-century idea that perception is “unconscious inference” is the root of his theory of the brain. Work in machine learning by Geoffrey Hinton and others on variational free energy provided the mathematics. Friston has repeatedly acknowledged W. Ross Ashby, especially the idea that a system that survives must regulate its essential variables and model its environment, as a direct ancestor of the free energy principle.

Contributions

Friston’s methods for analysing brain scans, including Statistical Parametric Mapping (SPM) and dynamic causal modelling, are among the most widely used tools in neuroscience.

His theoretical work, developed since the mid-2000s, is the free energy principle: any system that maintains itself against dissipation must, in effect, minimise the surprise of its sensory states, which it does by minimising a quantity called variational free energy. Active inference applies this to behaviour: organisms both update their models to fit their senses (perception) and act on the world to make their senses fit their models (action). The boundary between a system and its world is described as a Markov blanket.

Why it matters today

The free energy principle is one of the most ambitious current attempts at a unified theory of life and mind, and it is explicitly cybernetic: regulation, prediction and feedback in one formal framework. It is being applied in psychiatry, robotics and AI as an alternative to purely reward-driven learning.

Further reading

  • Karl Friston, “The Free-Energy Principle: A Unified Brain Theory?”, Nature Reviews Neuroscience (2010)
  • Thomas Parr, Giovanni Pezzulo and Karl Friston, Active Inference (2022)

Mentioned in