Skip to content

yazanobeidi/sentience

Repository files navigation

sentience

So far nobody has linked together the available building blocks for artificial general intelligence (AGI). These tools include Differentiable Neural Computers, Neural Programmer-Interpreters, Generative Adversarial Networks, and Deep Reinforcement Learning. Together, a single entity that is capable of ingesting multiple types of data, audio and visual recognition, general problem solving, learning to learn, algorithm generation, creativity, and interdisciplinary knowledge integration, may be created, resulting in something sort of like in the movie Her.

Sentience is deployed using Docker as a microservices for each distinct unit of the AGI:

  1. Executor

Executive functions and cognitive control. Appropriate behaviour selection and operation for goal achievement.

  1. Interactor

Active communication and interaction with others.

  1. Strategizer

Data aggregration, analysis, pattern recognition, forecasting, feeding to goal planning, definition of options, preferences, strategies, goals, contexts, and auxillary knowledge store.

An advantage of using Docker is that ensemble learning may be easily used between containers. Apache Kafka is used for messaging between microservices and is also run inside Docker.

Installation

Prerequisites:

a) Docker

b) Compose

b) Git

Steps:

  1. Clone the repository:

$ git clone https://github.com/yazanobeidi/sentience.git && cd sentience

  1. Start with build.sh (wraps docker-compose) :

$ ./build.sh --prod

  1. Say hello via the terminal!

Uninstall:

  1. Remove Docker images:

$ docker-compose down --rmi all

  1. Delete repository:

$ cd .. && rm -r sentience

Disclaimer

This project does not attempt to make the claim that strong AI is actually possible or not. What this project does is try to emulate intelligent behaviour using present state-of-the-art tools in the field.

Notes

python3.6 -m pip must be used to access pip for Python3.6, as python3.6-pip is not presently packaged with python3.6 in Ubuntu:trusty.

Author

Yazan Obeidi

Copyright

Yazan Obeidi, 2017-2018

Licence

Apache V2