References

active-control

@@ -0,0 +1,7 @@
+This document discusses the relationship between the frameworks of Active Inference (AIF) and Control as Inference (CAI) for adaptive action selection. 
+1. **Formalisms of AIF and CAI**: Both AIF and CAI can be formalized in the context of a partially observed Markov Decision Process (POMDP), where the goal is to infer the posterior distribution over latent variables like states and actions given observations.
+2. **Control as Inference**: CAI introduces an additional "optimality" variable to encode the notion of value, where the goal is to recover the posterior over states and actions given the belief that the agent will observe itself being optimal. The variational bound can be decomposed into terms related to extrinsic value, state divergence, action divergence, and observation ambiguity.
+3. **Active Inference**: Unlike CAI, AIF does not introduce additional variables to incorporate value, but instead assumes the generative model is intrinsically biased towards valuable states or observations. The expected free energy functional in AIF requires maximizing both extrinsic and intrinsic value.
+4. **Encoding Value**: The key distinction between AIF and CAI is how they encode value into the generative model, with AIF using biased priors and CAI using exogenous optimality likelihoods. This leads to differences in the objectives and exploratory behavior of the two frameworks.
+
+This document provides a formal comparison between the frameworks of Active Inference and Control as Inference, highlighting their similarities and key differences in how they encode the notion of "value" into their respective generative models.

intelligence-as-care

@@ -0,0 +1,7 @@
+This document presents a novel framework for understanding intelligence in terms of "Care", a metric focused on motivation, stress, and goal-directedness of agents. The key points discussed are:
+
+1. **Cognitive Light Cone Framework**: The scale of an agent's goals and the boundary of states it can possibly represent or work to change defines its "cognitive light cone", which can be used to compare diverse intelligences.
+2. **Two Distinct Light Cones**: Agents have a physical light cone (PLC) representing possible states, and a care light cone (CLC) representing their sphere of concern and motivation.
+3. **Bodhisattva Vow and Bodhisattva Path**: The Bodhisattva's commitment to universal compassion and infinite Care can be represented by an infinitely expanding CLC, suggesting it as a path to hyperintelligence.
+
+This framework integrates insights from biology, Buddhism, and AI to propose Care as a central invariant for understanding and advancing both natural and artificial intelligence.

perceptual-awareness

@@ -0,0 +1,13 @@
+This document is a research article that discusses the role of active inference in perceptual awareness. The key points are:
+
+**1. Active inference treats perception and action as inferential processes governed by minimizing free energy**: This variational perspective formalizes the notion of perception as hypothesis testing and treats actions as experiments designed to gather evidence for or against alternative hypotheses.
+
+**2. Troxler fading and binocular rivalry demonstrate the importance of active engagement with the sensorium**: Troxler fading, where peripheral percepts dissipate during fixation, and binocular rivalry, where awareness alternates between distinct percepts, can be understood as optimal inferences under the assumption that the world may change in an unpredictable way.
+
+**3. Attention can be understood as a covert action that resolves uncertainty**: Just as overt actions like eye movements are used to gather informative data, covert attentional shifts can serve a similar purpose, leading to perceptual alternations in the absence of overt movements.
+
+**4. The model makes testable predictions about the relationship between Troxler fading and binocular rivalry**: Individuals who report longer Troxler fading times should exhibit less belief updating during binocular rivalry, and pharmacological manipulation of catecholamine signaling should systematically affect Troxler fading.
+
+**5. The computational mechanisms underlying these phenomena can be mapped onto the functional anatomy of attention and perceptual awareness**: Reciprocal connections between sensory and motor regions, particularly in the right frontoparietal network, are likely to play a crucial role.
+
+This paper presents a unified computational account of Troxler fading and binocular rivalry, grounded in the principles of active inference. By emphasizing the role of action in perceptual awareness, it provides a novel perspective on these phenomena and generates testable hypotheses about their underlying neurobiology.