From a7c8df98a21b751bc9596a5006297d981d8297d4 Mon Sep 17 00:00:00 2001 From: Kostya Cholak Date: Sat, 28 Oct 2023 05:43:44 +0200 Subject: [PATCH] Add agi_definition.md --- docs/agi_definition.md | 18 ++++++++++++++++++ docs/world_model/knowledge_representation.md | 5 +---- 2 files changed, 19 insertions(+), 4 deletions(-) create mode 100644 docs/agi_definition.md diff --git a/docs/agi_definition.md b/docs/agi_definition.md new file mode 100644 index 0000000..1573435 --- /dev/null +++ b/docs/agi_definition.md @@ -0,0 +1,18 @@ +--- +title: AGI Definition +layout: default +nav_order: "25" +--- + +# Work In Progress + +# AGI Definition + +## Abstract Definition +On a high level we define AGI as a system that is capable of performing tasks in domains familiar to people with performance being close to people. +This abstract definition is centered around people for a good reason. There is no evidence that people are truly generally intelligent, we don't know if generation is binary, maybe there is a scale to it and people are just the most generally intelligent agents we are aware of. +So in order to not rely on assumption that general intelligence is binary we will set our goal to achieve the human level. + +## Technical Definition +AGI is a system that has sensory and control interfaces. Just like people have eyes, ears, etc and neurons to activate muscles. Sensors produce distorted raw image of the environment that is being processed with a mechanism for extracting abstractions. The system needs to transform raw data into abstract representations because environment is often too complex to work with it's raw representation. It would be probably impossible to think of the real world in terms of photons hitting our retina, so we use abstraction to represent the collections of those photons to think about real-world collections of elementary particles. After the system has abstract representation of the world it is capable to interact with the world in an intelligent way. There is no behavior that is intelligent objectively. Surviving isn't intelligent, it's just necessary for reproduction, which was necessary for our species to find itself alive today. So we won't set any requirements on the agents behavior that it must met in order to count intelligent. Instead we will define what functionality we need from it in order to be useful for people. +To be continued... \ No newline at end of file diff --git a/docs/world_model/knowledge_representation.md b/docs/world_model/knowledge_representation.md index e87d8e5..3bbbf8a 100644 --- a/docs/world_model/knowledge_representation.md +++ b/docs/world_model/knowledge_representation.md @@ -9,10 +9,7 @@ nav_order: "30" # Knowledge Representation ## Hierarchical Relation -Information like `apples are fruits` is stored in the knowledge base with `parents` relation between two concepts: +Information like "apples are fruits" is stored in the knowledge base with `parents` relation between two concepts: ![](../../assets/images/img1.png) This relation tells the agent that everything that is known about the concept `Fruit` is also applicable for the concept `Apple`. - - -