The Free Energy Principle: What might be the next biggest revolution in science

"Paraphrasing Einstein, everything should be made as simple as possible, but not simpler."

Take everything you know, everything you believe, and everything about life and existence. What does it all come down to? “Minimizing free energy.” That’s probably not the answer that was on your mind, but it is the answer that neuroscientist Karl Friston was thinking of. To be alive is to take what you expect to happen, whether it’s the weather tomorrow or your opponent’s next move in a game of chess, and what you see happening and make the difference between those two observations as small as possible. Paraphrasing Einstein, everything should be made as simple as possible, but not simpler. Minimizing free energy provides a way of achieving this.

Friston believes the free energy principle lies at the heart of all systems of life. The “free energy principle” gives a way of unifying all areas of research - from AI to biology to economics - behind the idea of “minimizing free energy.” Though it could be the next revolution in science, scientists and philosophers remain perplexed on what it means. How can we unify everything behind the concept of “minimizing free energy”? Do all systems really obey such a principle? What we do, what we see, and what we learn can all be unified through the free energy principle. This way, all systems remain in a type of equilibrium by exchange with their environment.

"Even down to the smallest protozoa, every organism wants to get to the point where they’re not too surprised by the world around them."

Using a Markov blanket, the set of useful information about a system that’s just enough to understand the world, you can then minimize the difference between what a model of the world predicts and the sense and perception associated with it. Even down to the smallest protozoa, every organism wants to get to the point where they’re not too surprised by the world around them. As the world changes, so does your perception of it. See the world, become surprised, and correct yourself accordingly. Repeat.

Take a look at how a caterpillar becomes a butterfly. Inside her chrysalis, protecting herself from the world, the caterpillar’s perception of the world changes through her metamorphosis. Moving away from her original food source and to the form of a pupa, the Markov blanket changes as the caterpillar decides what she needs from the world. As she becomes a butterfly, the caterpillar persists in the face of ongoing change, minimizing the free energy between what she sees and what she expects.

With Markov blankets using the free energy principle, computers may work the same way. Machine learning algorithms learn and adapt given new pieces of information. With these similarities, Friston has advised experts in artificial intelligence including computer engineers developing virtual assistant technologies, neuroscientists working on hearing aids, and startups that can predict diseases. The power of the free energy principle is expansive.

Written by Syed Hussain Ather. Illustrated by Melis Cakar.
Edited by Desislava Nesheva and Gil Torten.


Interested in supporting Knowing Neurons? Become a Patron today and help us deliver our mission of making neuroscience accessible to everyone.

Become a Patron!


Friston K. 2010. The free-energy principle: a unified brain theory? Nature Reviews Neuroscience. 10:127-138.

Friston K. 2013. Life as we know it. Journal of the Royal Society Interface. 355.

Schwartenbeck P., FitzGerald T., Dolan R. J., Friston K. 2013. Exploration, novelty, surprise, and free energy minimization. Frontiers in Psychology. 4.

Friston K., Kilner J., Harrison L. 2007. Free-energy and the brain. Synthese.

Schmidhuber J. 2010. Formal theory of creativity, fun, and intrinsic motivation (1990-2010). IEEE Transactions on Autonomous Mental Development. 2.

Avatar photo

Arielle Hogan

Arielle Hogan received a B.S. in Biology and a B.A. in French from the University of Virginia. She is now pursuing a Ph.D. in Neuroscience in the NSIDP program at UCLA. Her research focuses on CNS injury and neural repair. Specifically, she is researching the differential intrinsic transcriptional programs that allow for PNS regeneration and investigating how these transcriptional programs can be induced in models of CNS injury to promote regeneration. She also enjoys learning about biomechatronics and brain-machine interface (BMI), as well as participating in science outreach and teaching. Outside of the lab, she spends time practicing her French, playing basketball, watching movies (even the bad ones), and traveling. For more information about Arielle Hogan, please visit her full profile.