Freud and the Algorithm: Neuropsychoanalysis as a Framework to Understand Artificial General Intelli
Citation Information
Author: Luca M. Possati
Title: Freud and the Algorithm: Neuropsychoanalysis as a Framework to Understand Artificial General Intelligence
Publisher: Humanities and Social Sciences Communications
Publication Year: 2021
Affiliation: University of Porto
Abstract
This paper proposes that neuropsychoanalysis, emphasizing the role of emotions and affect in cognition, provides an innovative paradigm for Artificial General Intelligence (AGI). In contrast to the cortical-centric view of current AGI research, which predominantly focuses on higher-order cognitive functions, the author suggests a subcortical model based on Jaak Panksepp’s primary affective systems. This approach advocates for incorporating emotions as intrinsic functions of AGI systems to better simulate human-like intelligence and control over autonomous processes. By grounding AGI in neuropsychoanalysis, it may become feasible to address complex issues in AGI, including AI control.
Keywords
Key Findings and Analysis
Introduction
The paper opens by establishing the need for an alternative approach to AGI development that emphasizes affective neuroscience and neuropsychoanalysis over conventional cognitive science frameworks. Current AGI research focuses on the brain's cerebral cortex and high-level cognitive functions, which does not sufficiently address the role of subcortical structures associated with emotions and instincts in AGI development. The author proposes that a neuropsychoanalytic model, particularly one informed by the works of Freud, Jaak Panksepp, and Mark Solms, could more accurately simulate the affective and instinctual aspects of human intelligence.
Objectives
The objective is to define and explore principles for AGI design that prioritize subcortical functions and integrate Panksepp's seven affective systems—SEEKING, RAGE, FEAR, LUST, CARE, PANIC, and PLAY—as foundational elements of AGI.
Neuropsychoanalysis as an Alternative Framework
Neuropsychoanalysis combines insights from psychoanalysis and neuroscience to explore the dual-aspect nature of the brain/mind relationship. According to neuropsychoanalysis, mental and physical processes are interconnected but are experienced through two distinct but parallel forms: subjective first-person (mind) and objective third-person (brain) perspectives. By integrating both views, neuropsychoanalysis avoids reducing complex mental phenomena to biochemical processes alone.
Key Points in Neuropsychoanalysis
Dual-aspect Monism: This concept underlies neuropsychoanalysis, positing that mind and brain are two perceptions of a single reality, thus demanding integrated approaches for understanding AGI.
Subjective and Objective Knowledge Integration: Neuropsychoanalysis validates both subjective psychological experiences and objective neuroscientific data.
Psychoanalytic and Evolutionary Basis: The emphasis is on foundational instincts, such as fear and desire, as essential components for AGI.
Revisiting the Affective Systems of Panksepp
Panksepp’s affective neuroscience model identifies primary emotional systems rooted in the subcortical brain. These systems operate independently of cognitive processing and offer essential survival functions. Each system, like SEEKING or FEAR, contributes to behavioral drives and responses crucial for interacting with environmental stimuli.
Primary Affective Systems for AGI
SEEKING: Drives motivation and exploration, forming a basis for curiosity and goal-driven behaviors.
RAGE: Enables self-preservation by responding to threats and frustration.
FEAR: Promotes survival through threat avoidance.
LUST: Manages reproductive behaviors and social bonding.
CARE and PLAY: Facilitate social bonds and cooperative behavior.
Challenges with Traditional AI and AGI Models
Traditional AGI approaches focus on cognitive tasks like language, memory, and problem-solving but neglect emotional and instinctual processes. This cortical-centric view is limited, as it overlooks the subcortical systems that play critical roles in real-world decision-making, motivation, and social interaction.
Limitations of Purely Cognitive Models
Lack of Emotional Modulation: Current models cannot accurately replicate emotions' effects on decision-making and learning.
Absence of Contextual Reactions: Emotions like fear and rage dynamically adapt behaviors to contextual threats, which purely logical models cannot simulate.
Inability to Form Self-regulating Homeostasis: Without a subcortical foundation, AGI systems struggle with the adaptive regulation of physiological and emotional states.
A New Approach for AGI Design: Integration of Markov Blankets and Causal Inference
The paper suggests using a Markov Blanket framework to structure AGI, which would demarcate internal and external states, simulating self-organizing boundaries that mimic biological systems' autonomy. This model emphasizes maintaining a balance between internal states and environmental interactions. Possati also integrates Judea Pearl’s causal analysis framework, advocating a shift from purely probabilistic models to those recognizing causal relationships, which is essential for building an AGI system capable of adaptive learning and genuine interaction with its environment.
Key Components of the AGI Framework
Active Inference and Free Energy: Systems should minimize "free energy" to predict and adjust responses to external stimuli.
Dynamic Causality: Unlike Bayesian networks that merely predict associations, Pearl’s causal model would enable AGI to understand and respond to causative factors, reflecting more accurate human-like reasoning.
Practical Application: Affective-Centered Algorithms
The author proposes that algorithms designed for AGI should reflect the neuropsychoanalytic structure by incorporating Panksepp’s affective systems. For example:
Modeling SEEKING through Dopaminergic Systems: Motivates exploration and learning.
Implementing a RAGE System: Manages threat responses, which could enhance resilience in autonomous systems.
Conclusion and Implications for Future AGI Research
Possati concludes that neuropsychoanalysis provides a paradigm shift for AGI by establishing the groundwork for emotionally intelligent systems. He argues that a subcortical approach to AGI, grounded in affective neuroscience, will improve AGI’s adaptive capabilities and better align it with human emotional and motivational structures. This approach may also address ethical concerns related to AGI control, as a system rooted in affective processes may allow more natural regulatory mechanisms than purely logic-driven architectures.
This paper provides an interdisciplinary framework, blending psychoanalytic theory with computational models, to redefine AGI's foundations by rooting it in affective neuroscience rather than exclusively cognitive paradigms.
Last updated