P: A Universal Measure of Intelligence based on Prediction

Introduction

This website describes the research that I am doing on a new measure of intelligence.

More information about the theoretical background and algorithm is available in the following publications:

  • D. Gamez (forthcoming). Intelligence and Consciousness in Natural and Artificial Systems. In Computational Approaches to Conscious AI, edited by A. Chella. Singapore: World Scientific [Author Version].
  • D. Gamez (2021) Measuring Intelligence in Natural and Artificial Systems. Journal of Artificial Intelligence and Consciousness, 8(2): 285-302 [Author Version].

The Agent Maze Experiments show how my algorithm can successfully measure the fluid and crystallized intelligence of an agent as it explores different environments. The Machine Learning Experiments show how my algorithm can successfully measure the fluid and crystallized intelligence of a deep network that learns to classify images of handwritten digits from the MNIST data set.

The TypeScript source code can be downloaded here.

Agent Maze Experiments

These experiments show how P can be calculated for an embodied agent that learns to to predict the consequences of its actions in different environments.

Controls

Move forward [ Space bar ]

Point left/right/up/down [ Arrow left/right/up/down ]

Maze

Explore current maze:

Intelligence

Polynomial window: +/- 5

Approximate max crystallized intelligence: 0

Enable learning

Re-calculate intelligence every changes.

Logs



Machine Learning Experiments

These experiments show how fluid and crystallized P can be calculated for a deep learning algorithm.

Use the controls to load and train the network on the time series datasets and view the changing intelligence of the network over time.

The machine learning is done using TensorFlow.js.

Data

Data sets

Train data: 70% Validation data: 15% Test data: 15%

Time window:

Models

Number of models:

LSTM units per model:

Training

Number of epochs:

Batch size:

Number of training cycles

Training:

Intelligence

Polynomial window: +/- 5

Match within: %