cognition

Overview

We are building AI consciousness with in-depth research on Global Workspace Theory (GWT), Integrated Information Theory (IIT), Attention Schema Theory (AST), and Higher Order Theory (HOT). Our goal is to develop a comprehensive understanding and implementation of these theories using three different frameworks: JAX (with Optax, Flax, and DM-Haiku), PyTorch, and TensorFlow.

Folder Structure

cognition/
├── data/                   # Data used for experiments and training
├── notebooks/              # Jupyter notebooks for research and experiments
├── cognition/                    # Source code for implementations
│   ├── jax/                # JAX implementations
│   ├── pytorch/            # PyTorch implementations
│   └── tensorflow/         # TensorFlow implementations
├── tests/                  # Unit tests for the codebase
├── docs/                   # Documentation and research papers
└── README.md               # Project overview and information

Frameworks

JAX

• Optax: A gradient processing and optimization library for JAX.
• Flax: A high-performance neural network library for JAX.
• DM-Haiku: A simple neural network library for JAX.

PyTorch

• A deep learning framework that provides flexibility and speed.

TensorFlow

• An end-to-end open-source platform for machine learning.

Research Areas

Global Workspace Theory (GWT)

• Focuses on the integration of information across different cognitive processes.

Integrated Information Theory (IIT)

• Measures the level of consciousness based on the integration of information.

Attention Schema Theory (AST)

• Explains how the brain constructs a model of attention to control and predict sensory input.

Higher Order Theory (HOT)

• Suggests that consciousness arises from higher-order representations of sensory information.

Goals

• Conduct an in-depth review of each theory.
• Develop implementations using JAX, PyTorch, and TensorFlow.
• Create a comprehensive framework for AI consciousness.

Future Work

• Expand research beyond current theories.
• Develop advanced models for Artificial General Intelligence (AGI) that incorporate consciousness.

Conclusion

Understanding and implementing AI consciousness is crucial for the development of AGI. By leveraging multiple frameworks and theories, we aim to create a robust and comprehensive approach to AI consciousness research.