This project is a 👽system, it doesn't work like your usual machine learning (ML) systems.
It is a agent with true autonomy 🤖🧠, it deals with situations in realtime (<200ms reaction time), it can run all day, it learns lifelong and incrementally.
This project consists out of the intelligent "core" and comes with a vision subsystem. A subsystem to communicate with the system in natural language is also included. This repository doesn't come pre-packaged with other tooling/subsystems.
The vision subsystem (which isn't integrated/used yet) also works in realtime, it also learns lifelong and incrementally. See in https://github.com/PtrMan/23R/tree/main/moduleVision for the vision system.
curl https://nim-lang.org/choosenim/init.sh -sSf | sh
There are two ways to run, the shell is preferable to do experiments with the reasoner, while the web interface is much more user friendly, but the reasoner there runs in real-time and always processes requests as soon as they arive.
shell
nim compile --run entryShell.nim
Note that the shell waits for user input and doesn't run in real-time because it waits for input!
web mediator+net-NAR+webserver
start mediator
python ./partMediator/mediator.py &
run webserver
node ./partUiServer/a.js &
run net-NAR
nim compile --threads:on --run entryNetUdpClient.nim
write > as a prefix in the shell. This is currenlty not implemented in the web interface!
example:
>Tom is a furry cat
The formal language to communicate with the system with non-natural language is done in narsese.
This system is a implementation of a non-axiomatic reasoning system. It comes with some additions on top of the NARS-theory which are inspired by relational-frame theory.
The logic follows non-axiomatic logic, which is described in serveral publications such as (1).
Wait! Why doesn't this come pre-packaged with 'batteries included' like Python or a webbrowser etc.?
To keep this project/repository simple as in keep-it-small-stupid (KISS).
Technical documentation (draft) link