The future of project-origin

This project is a seed for a far-reaching vision. In the future, it should allow developing a wide variety of universes. Below are some examples of future capabilities of project-origin:

• project-origin should allow extending the physics of the environment easily and maybe incorporate realistic physical simulators.
• It should support morphology of creatures, i.e. to give physical shape to creatures, that may affect their capabilities such as movement speed, power, etc.
• It should support the introduction of different inanimated and non-intelligent living objects with a variety of functionalities and behavior to the universe, e.g. poison.
• While currently, space is a 2-D, in the future it should allow supporting different types of spaces.
• Easily controlling biological aspects of physics such as mating rule, evolution, and intelligence inheritance.
• It should allow extending the creature’s capabilities, such as adding vocal communication and even love, hate, and motivation.
• It should allow defining dynamic natures, like periods of dearth and epidemics.
• Use dynamic graph deep learning framework such as TensorFlow 2.0 or Pytorch.
• Use openAI baselines and/or Google dopamine projects as an implementation of the creature brain.
• Using Unity or other engines, implement a visual simulation environment.
• Add morphology and form to creatures that define its biological and physical features.

Far-reaching plans and directions

• Model selection and hyper-parameter tuning. Show that Convolutional can better survive than fully connected layers.
• Special relation between individuals based on graph networks to process the relationship between individuals.
• Hebbian Learning vs. Gradient-based in a survival environment. Which is better?
• Develop a game platform which allow training of a race and then allow it compete over resources with different race.
• compare between the different implementation of RL algorithms, model-based and model-free, policy gradient and TD learning.