free energy principle
Karl Friston's proposed unifying theory that the brain works to minimize prediction error (surprise) about sensory inputs.
As defined in ms_54
Other definitions
Karl Friston's theory that intelligence minimizes surprisal (expected surprise/entropy) by building predictive models of the environment, which allows systems to extract and use free energy more efficiently
Forget Everything You Believed About Computing w/ Gill Verdon | EP #102
Where it comes up · 17
- 320 - Constructing Self And World3 claims
- Sholto Douglas & Trenton Bricken — How LLMs actually think2 claims
- 235 | Andy Clark on the Extended and Predictive Mind2 claims
- AMA | August 20212 claims
- Michael Levin - Why Intelligence Isn't Limited To Brains.1 claim
- Philip Ball1 claim
- 478 - The Psychedelic Mind1 claim
- 100 | Solo | Life and Its Meaning1 claim
and 9 more
Most important claims · 10 of 22Search all claims for “free energy principle” →
- Even if the correct action is expressible as optimizing a number, humans never actually do this perfect optimization because we are bounded in both collecting and processing information; in practice we always satisfice (settle for good enough), as reflected in approaches like Friston's free energy principle that approximate Bayesian updating quickly.AMA | February 2023
- Entropy is not a relational phenomenon; for a given system with a given coarse-graining, entropy is simply high or low, full stop. What it requires is a coarse-graining — a chosen way of grouping microstates into indistinguishable macrostates — which essentially everyone agrees on but which we nonetheless must choose; once chosen, entropy is just a calculable number not related to anything else.AMA | October 2022
- Behavioral/evolutionary simulations show that if agents must perform more than a certain number of tasks with basic tools and objects, they learn a ground-truth representation of the world rather than cheap visual heuristics—because the many possible use cases require knowing what an object actually is—suggesting LLMs trained on diverse human data learn genuine features of the world, making them potentially less alien.Sholto Douglas & Trenton Bricken — How LLMs actually think
- Karl Friston's free energy principle holds that the brain builds a model of the world to minimize surprise, but the apparent counterexample of deliberately seeking surprise is resolved by noting that complex organisms minimize net surprise over the future; however, this cannot be the whole story, because one could minimize surprise by locking oneself in a dark room—the real constraints come from homeostatic needs (hunger, thirst, reproduction, social approval) working at cross purposes.100 | Solo | Life and Its Meaning
- Perception and action are two ways of getting rid of prediction error using the same basic neuronal operations: perception minimizes error by finding a better prediction or retrieving an old model of the world to fit the sensory flow, while action minimizes error by changing the world to fit the prediction.235 | Andy Clark on the Extended and Predictive Mind
- Per the free energy principle, living things must not only predict their environment but also predict themselves; organisms are patchworks at different scales of systems that cooperate and compete, each guessing what will happen next and having preferences, all working to minimize the delta between expectation and outcome, which is critical for evolvability.132 | Michael Levin on Growth, Form, Information, and the Self
- Friston's free energy minimization is a higher-road version of the prediction story: rather than starting from perception and action, it starts from what it takes to be a persistent living system, which must preferentially inhabit the states that define it, so a fish in a broad sense predicts it ought to be in water because its actions are all designed with that in mind.235 | Andy Clark on the Extended and Predictive Mind
- Minimizing prediction error alone is insufficient; the free energy principle requires minimizing both the prediction error against sensory data and the informational distance between the generative model's output and prior probabilities, so perception balances sensory evidence against prior expectations.320 - Constructing Self And World
- Although the free energy principle and Bayesian brain framework can be read as saying brains minimize surprise (which would imply sitting in a dark room doing nothing), the correct reading is that we minimize net surprise over our whole lives, which requires exploring and doing weird things now so we can better anticipate what's coming—mirroring the explore/exploit trade-off and avoiding the failure case of doing the same thing repeatedly.308 | Alison Gopnik on Children, AI, and Modes of Thinking
- The old consensus that the brain takes bottom-up sensory data and processes it up a stack until it figures out what is happening is precisely upside down; the latest theory holds the brain generates a top-down simulation and only sends prediction errors upward.320 - Constructing Self And World