This is a research project about General Artificial Intelligence system loosely based on Practopoiesis Theory which stands for neural network that depends purely on its environment and instead of converting inputs into symbols that are processed by "machine" to give output, it processes signals as energy potentials and by evolution of connections and constant change of brain structure, it produces "consciousness" naturally.
You can read more about Practopoiesis Theory here: http://www.danko-nikolic.com/practopoiesis/
All toolset modules are written in Rust programming language (a multi-paradigm systems programming language focused on safety, especially safe concurrency) and is available on crates.io as Rust Crate ready to be your project dependency.
Psyche toolset provides FFI libraries and wrappers for many languages and frameworks.
Every blue line is a connection between two neurons and every white dot is a signal traveling through neural network.
Each spore has its own brain connected to body sensors (smell) and motors (legs) and by that it tries to find and eat food portions left in water. You can also manipulate environment by producing fluid currents with mouse dragging.
Record in Cargo.toml
:
[dependencies]
psyche = "0.2"
Your crate module:
extern crate psyche;
use psyche::core::brain_builder::BrainBuilder;
use psyche::core::config::Config;
use psyche::core::Scalar;
// prepare config for brain.
let mut config = Config::default();
config.propagation_speed = 50.0;
config.synapse_reconnection_range = Some(15.0);
config.neuron_potential_decay = 0.1;
config.synapse_propagation_decay = 0.01;
config.synapse_new_connection_receptors = Some(2.0);
// build brain.
let mut brain = BrainBuilder::new()
.config(config)
.neurons(100)
.connections(200)
.min_neurogenesis_range(5.0)
.max_neurogenesis_range(15.0)
.radius(30.0)
.sensors(10)
.effectors(10)
.brain();
loop {
// trigger sensors.
for sensor in brain.get_sensors() {
brain.sensor_trigger_impulse(sensor, 1.0);
}
// process brain step.
brain.process(1.0);
// read effectors and act based on their stored potential.
for effector in brain.get_effectors() {
if let Ok(potential) = brain.effector_potential_release(effector) {
println!("{:?} = {:?}", effector, potential);
}
}
}