Gradients in neural networks explain how experience feels over time
Gradland: On Phenomenal Experience, Differentiated Across Many Dimensions
Artificial IntelligenceNeural and Evolutionary Computing
Summary
This paper looks at how the structure of physical interactions inside neural networks relates to the way we experience the world. The authors study an idealized environment called Gradland, where these networks follow smooth, detailed rules allowing measurement of how their internal connections change. They use two new ways to measure these changes and find that these measures help explain things like how long experiences last, why some feelings are clear and others vague, how textures feel, what newborns might experience, and how learning feels. The work also explores why having rich and detailed experiences is useful.
What this means in practice
- •For ai developers: Design neural network models that better capture how experience changes in clarity and duration by using gradient-based structural measures.
- •For cognitive systems engineers: Create systems that simulate naturalistic sensory experiences by linking Jacobian structure to perception features like texture and clarity.
Tested on simulated data.
Authors
David Balduzzi
Abstract
This paper investigates the hypothesis that the first-order structure of physical interactions, i.e. gradients or Jacobians, characterizes the structure of phenomenal experience. It does so in an idealized world inhabited by neural networks, Gradland, where the physics are known and the functions are (mostly) differentiable. The paper introduces two measures of Jacobian structure: effective rank and cohesion, based on Kirchhoff complexity. Applying the measures to a series of worked examples shows the hypothesis accounts for: (1) the duration of experience, that it can prolong over hundreds of milliseconds; (2) the difference between what is experienced vividly and obscurely; (3) the experience of texture; (4) the blooming buzzing confusion presumably experienced by newborns; (5) the difference between ideas that are held distinctly in mind and ideas that are confused; (6) what learning is like; and finally (7) the paper explains the function of rich, dense experience.