Uploaded June 2026 | Updated November 2026, 1 day ago
Book: Fundamentals of Active Inference
Principles, Algorithms, and Applications of the Free Energy Principle for Engineers, by Sanjeev V. Namjoshi https://mitpress.mit.edu/9780262050951/fundamentals-of-active-inference/
Register: http://textbook-group.activeinference.institute/
Code: github.com/ActiveInferenceInstitute/fundamentals
Playlist of Textbook Group: youtube.com/playlist?list=PLNm0u2n1IwdrpoTg-9l37YIBiDrrOzTw-
Direct link to group document: coda.io/d/_dZHp_UGhVZE/Meetings_su3ToqMp
------
CHAPTERS
0:00 Intro: Part One is finished
1:25 Should the pace change for Part Two?
4:09 A recommended stream on predictive coding
5:31 LaTeX formatting for the notes
8:19 What Part One left out: action, and anything dynamic
9:41 Part Two: generative processes that change over time
11:06 Recap: generative process vs generative model
12:27 Counterfactual predictions and model evidence
13:50 Chapter 3 as Chapter 2 redux: setting the dials
15:15 The parameter cost surface and gradient descent
16:38 Why hierarchical models: processes at different rates
18:04 Chapter 3's culminating algorithm
19:28 Chapter 4: approximating exact Bayesian inference
20:53 Surprisal and the four canonical forms of free energy
22:18 Chapter 5: re-expressing free energy as prediction error
23:40 How the chapters connect
25:02 Q&A: where noise enters in perception-only models
26:28 The linearity assumption in Part One
27:51 Food size and light intensity as a linear relationship
29:16 Gaussians, uniforms, and what the likelihood assumes
30:43 Where uncertainty about the mapping is encoded
32:06 The equation of a line, and what else you could plug in
33:33 Q&A: a thought experiment about groups and shared models
34:59 Hunter-gatherer bands as distributed information processing
36:22 What is the unit of analysis for an informational process?
37:01 Individual and group-level generative models
38:07 Where the book does and does not address multi-agent agency
39:05 Compositionality as the open question
40:28 Generative models as attractors in human systems
41:50 Building models from simple pieces: what to include
43:13 What variables explain a person interacting with a community?
44:37 Epistemic communities under active inference: the simulation
46:00 Why polarisation follows from minimising uncertainty
47:26 Communities converging on their own truth
48:51 Q&A: people infer whether to believe the messenger
49:57 Precision, relevance and what to attend to
50:14 Why the book stays with the single agent
50:43 Correcting active inference as a verb
53:08 Where to put questions asynchronously
54:32 Q&A: is this a correct summary of the active inference process?
56:29 Inference and parameter learning as two problems
58:13 Q&A: is gradient descent used in perception too?
59:32 Q&A: relationship between active inference and reinforcement learning
1:00:52 Information gain as the key difference
1:01:45 Reward-driven agents vs curiosity
1:04:02 Epsilon-greedy vs a principled account of exploration
1:05:34 Interpretability as a goal, not just performance
1:06:20 Q&A: can an agent quantify its own uncertainty?
1:08:01 Reasoning about distributions from the very beginning
1:08:48 Uncertainty about things that have not happened yet
1:09:24 Entropy, precision, and the volume-knob analogy
1:11:28 Q&A: turning precision up so the model outweighs the eyes
1:12:04 Why priors matter at all
1:12:29 Waking in the dark: what a prior stabilises
1:13:35 Why this matters for psychiatry and communication
1:14:07 Figure 2.11: the effect of prior precision on belief updating
1:14:41 Wrap-up/_dZHp_UGhVZE/Meetings_su3ToqMp
------
Active Inference Institute information:
Website: https://activeinference.institute/
X: https://x.com/InferenceActive
Discord: https://discord.activeinference.institute/
Donate: https://donate.activeinference.institute/
YouTube: youtube.com/c/ActiveInference
Active Inference Livestreams: https://video.activeinference.institute/
Book: Fundamentals of Active Inference
Principles, Algorithms, and Applications of the Free Energy Principle for Engineers, by Sanjeev V. Namjoshi https://mitpress.mit.edu/9780262050951/fundamentals-of-active-inference/
Register: http://textbook-group.activeinference.institute/
Code: github.com/ActiveInferenceInstitute/fundamentals
Playlist of Textbook Group: youtube.com/playlist?list=PLNm0u2n1IwdrpoTg-9l37YIBiDrrOzTw-
Direct link to group document: coda.io/d/_dZHp_UGhVZE/Meetings_su3ToqMp
------
CHAPTERS
0:00 Intro: Part One is finished
1:25 Should the pace change for Part Two?
4:09 A recommended stream on predictive coding
5:31 LaTeX formatting for the notes
8:19 What Part One left out: action, and anything dynamic
9:41 Part Two: generative processes that change over time
11:06 Recap: generative process vs generative model
12:27 Counterfactual predictions and model evidence
13:50 Chapter 3 as Chapter 2 redux: setting the dials
15:15 The parameter cost surface and gradient descent
16:38 Why hierarchical models: processes at different rates
18:04 Chapter 3's culminating algorithm
19:28 Chapter 4: approximating exact Bayesian inference
20:53 Surprisal and the four canonical forms of free energy
22:18 Chapter 5: re-expressing free energy as prediction error
23:40 How the chapters connect
25:02 Q&A: where noise enters in perception-only models
26:28 The linearity assumption in Part One
27:51 Food size and light intensity as a linear relationship
29:16 Gaussians, uniforms, and what the likelihood assumes
30:43 Where uncertainty about the mapping is encoded
32:06 The equation of a line, and what else you could plug in
33:33 Q&A: a thought experiment about groups and shared models
34:59 Hunter-gatherer bands as distributed information processing
36:22 What is the unit of analysis for an informational process?
37:01 Individual and group-level generative models
38:07 Where the book does and does not address multi-agent agency
39:05 Compositionality as the open question
40:28 Generative models as attractors in human systems
41:50 Building models from simple pieces: what to include
43:13 What variables explain a person interacting with a community?
44:37 Epistemic communities under active inference: the simulation
46:00 Why polarisation follows from minimising uncertainty
47:26 Communities converging on their own truth
48:51 Q&A: people infer whether to believe the messenger
49:57 Precision, relevance and what to attend to
50:14 Why the book stays with the single agent
50:43 Correcting active inference as a verb
53:08 Where to put questions asynchronously
54:32 Q&A: is this a correct summary of the active inference process?
56:29 Inference and parameter learning as two problems
58:13 Q&A: is gradient descent used in perception too?
59:32 Q&A: relationship between active inference and reinforcement learning
1:00:52 Information gain as the key difference
1:01:45 Reward-driven agents vs curiosity
1:04:02 Epsilon-greedy vs a principled account of exploration
1:05:34 Interpretability as a goal, not just performance
1:06:20 Q&A: can an agent quantify its own uncertainty?
1:08:01 Reasoning about distributions from the very beginning
1:08:48 Uncertainty about things that have not happened yet
1:09:24 Entropy, precision, and the volume-knob analogy
1:11:28 Q&A: turning precision up so the model outweighs the eyes
1:12:04 Why priors matter at all
1:12:29 Waking in the dark: what a prior stabilises
1:13:35 Why this matters for psychiatry and communication
1:14:07 Figure 2.11: the effect of prior precision on belief updating
1:14:41 Wrap-up/_dZHp_UGhVZE/Meetings_su3ToqMp
------
Active Inference Institute information:
Website: https://activeinference.institute/
X: https://x.com/InferenceActive
Discord: https://discord.activeinference.institute/
Donate: https://donate.activeinference.institute/
YouTube: youtube.com/c/ActiveInference
Active Inference Livestreams: https://video.activeinference.institute/

![ActInf GuestStream 119.1 Thermal Macroeconomics: An axiomatic theory of aggregate economic phenomena
Thermal Macroeconomics: An axiomatic theory of aggregate economic phenomena
Nick Chater (Warwick Business School) and Robert MacKay (Mathematics Institute, University of Warwick) (joint first authors)
Thermal Macroeconomics: An axiomatic theory of aggregate economic phenomena
N.J.Chater, R.S.MacKay
https://arxiv.org/abs/2412.00886
arXiv:2412.00886 [econ.GN]
An axiomatic approach to macroeconomics based on the mathematical structure of thermodynamics is presented. It deduces relations between aggregate properties of an economy, concerning quantities and flows of goods and money, prices and the value of money, without any recourse to microeconomic foundations about the preferences and actions of individual economic agents. The approach has three important payoffs. 1) it provides a new and solid foundation for aspects of standard macroeconomic theory such as the existence of market prices, the value of money, the meaning of inflation, the symmetry and negative-definiteness of the macro-Slutsky matrix, and the Le Chatelier-Samuelson principle, without relying on implausibly strong rationality assumptions over individual microeconomic agents. 2) the approach generates new results, including implications for money flow and trade when two or more economies are put in contact, in terms of new concepts such as economic entropy, economic temperature, goods values and money capacity. Some of these are related to standard economic concepts (eg marginal utility of money, market prices). Yet our approach derives them at a purely macroeconomic level and gives them a meaning independent of usual restrictions. Others of the concepts, such as economic entropy and temperature, have no direct counterparts in standard economics, but they have important economic interpretations and implications, as aggregate utility and the inverse marginal aggregate utility of money, respectively. 3) this analysis promises to open up new frontiers in macroeconomics by building a bridge to ideas from non-equilibrium thermodynamics. More broadly, we hope that the economic analogue of entropy (governing the possible transitions between states of economic systems) may prove to be as fruitful for the social sciences as entropy has been in the natural sciences.
[Submitted on 13 Jun 2025]
Convergence to equilibrium for a class of exchange economies
R.S.MacKay
https://www.arxiv.org/abs/2506.11770
arXiv:2506.11770 [math.PR]
For a class of stochastic dynamical models of exchange economies that we call ``fully connected Cobb-Douglas, the paper proves convergence of the probability distribution to an equilibrium, in total variation metric as time goes to infinity. The convergence is exponential and the equilibrium is determined uniquely by the number of agents, their ``exponents, and the initial amounts of money and goods in the economy.
Active Inference Institute information:
Website: https://www.activeinference.institute/
Activities: https://activities.activeinference.institute/
Discord: https://discord.activeinference.institute/
Donate: http://donate.activeinference.institute/
YouTube: https://www.youtube.com/c/ActiveInference/
X: https://x.com/InferenceActive
Active Inference Livestreams: https://video.activeinference.institute/ ActInf GuestStream 119.1 Thermal Macroeconomics: An axiomatic theory of aggregate economic phenomena](https://i.ytimg.com/vi/O4JkF40yvqU/mqdefault.jpg)








