NARS is a universal reasoning system, like a "universal Turing Machine" that can emulate any Turing Machine, and serves as a formal model of an intelligent system

following same principles as the human mind and able to solve problems in various domains.

NARS is designed in a layered manner. In its lowest level, NARS can be seen as a Turing machine with limited resources and with algorithms implemented to use effectively NAL (Non-axiomatic logic - which is also layered). It reasons (solve problems in a case by case manner) and learns (from its own experience) on the same time. NARS is a superset of Turing machine, and while it has been not mathematically proved yet, it is also able to emulate a Turing machine or any other computational or logic system.

Although NARS uses for reasoning fixed algorithms, the process might not be repeatable depending on many factors, such as amount of believe the system currently have, availability of resources, amount of task currently being processed etc. So even though NARS uses deterministic algorithms, its outcome is not always predictable.

In its highest level, NARS can reason about itself (NAL-9 which together with NAL-8 represents systems’ self-awareness and self-control).

NARS can work with information of any level of abstraction, can reason, learn, and derive new information from existing one. NARS tolerates incompleteness and incorrectness, however avoids generating “nonsenses” (system requires, that information entering to the inference rule, must be relevant).

NARS have been designed with simplicity and consistency in mind, it should be possible to implement in SW or HW in a very effective way.

It is a strong believe, that NARS is “AGI complete” (similar, to Turing completeness in computing system).

NARS promises very broad problem solving capabilities, which as it is with Turing machine, might vary on its real implementation. Despite the fact, that the resource limitation is already in its definition (AIKR).

The real strength of the system is in situation when the conventional techniques fails.


Background Image

Header Color


Content Color