Our last public lecture for this year’s series on Biological Complexity at the Royal Institution was delivered by Karl Friston, Wellcome Centre for Neuroimaging, UCL. Karl has done seminal work in theoretical neuroscience and he is currently working on Bayesian models for the brain. He took us through a fascinating journey on what it means to be alive, and what conditions have to be satisfied for defining a living thing. Then Markov blanket was introduced and explained very clearly with an eye to defining what counts as a dynamical system able to generate recursive patterns over a period of time. From the statistical techniques, the discussion soon moved to the brain and the principle of minimising prediction-error as a key mechanism through which our brain builds a representation of the outside world, anticipates problems and challenges and adjusts sensory and action to respond to those problems. Simulated bird songs ended the talk as an illustration of the basic mechanism of prediction-error minimisation, with an enchanted audience that was reminded of entropy, Plato’s cave, and Helmholtz’s work on acoustics. Integration of philosophy and the sciences at its best, with a clear Kantian underlying tone in the idea that the world does not come to us as given but our brain builds it via feed-forward mechanisms of prediction and error.