Skip to content
/ 20NAR1 Public

General Machine Intelligence (GMI) Project Inspired by NARS

License

Notifications You must be signed in to change notification settings

PtrMan/20NAR1

Repository files navigation

This is a NARS inspired General Machine Intelligence (GMI) system. The current implementation only implements a NAR(non-axiomatic reasoner) for NAL1-6 and NAL7-8. NAL-9 is currently only supported with a special operator.

working definition of intelligence:
solving complex and complicated problems under adaptation with insufficient knowledge and resources
is a composition of the definition of intelligence from Dr. Pei Wang paper and the definition of intelligence from Dr. Ben Goertzel.

how to run

Running the program is easy. Note that cargo test can be omitted if the unittests should be skipped.

how to run in interactive mode

cargo test && cargo run --release it

how to run in server mode

cargo test && cargo run --release srv
connect to server with for example netcat: nc 127.0.0.1 2039

how to run pong

cargo test && cargo run --release envPong3

how to eval Q&A performance for one file

cargo test && cargo run --release bQA nalExp/ExTimAnimalProd0.nal

build documentation

cargo doc --lib

how to train the NN for NLP

rm -f ./logX.txt && cargo test && cargo run --release nnTrain.nlp

Documentation

Documentation will follow when the features(mechanisms) and implementation has settled to a "stable" state.
A paper describing novel parts will follow.
The API documentation can be created from the source code by building the documentation.

commands

The system accepts either Narsese or commands as inputs. Commands are used to give the reasoner compute in the form of cycles. Commands also can be used to change parameters of the input/output and/or the reasoner itself. Commands also allow to invoke special functionality of the NARS+ implementation, for example for the NLP module.
See in source file NarInputFacade.rs for a list of all implemented commands.

dependencies

basic

  • rust (minimum 1.36, latest stable recommended)

when using python binding

  • python 3.X, recommended 3.5 and up

robotics

Robot examples need pybullet, install with pip install pybullet

what is implemented

temporal

  • unification of =/> by subj
  • goals
  • decision making
  • Q&A for temporal, ex: <(a,^x) =/> b>?

non-temporal

  • Q&A

what is missing?

  • unification of =/> by pred
  • use of non-temporal knowledge for temporal inference halfdone
  • quests