Building an Open Artificial General Intelligence
-
An agent interacting with the world [modeled as an partially observable (only seen through the sensors) environment]. 2.1 Abilities: 2.2. Move - expressively, purposefully, naturally 2.3. See - people, objects, motion, expressions, gestures 2.4. Speak - in a variety of voices and languages 2.5. Listen - to speech, music, and other sounds 2.6. React and Interact - physically and verbally 2.7. Teach, Learn, Behave, Intend, Consider, Emote, and Remember
-
World means very everthing (physical and none physical) in the world also includes other agents (mutli-agent environment all having their own objective functions). Objective functions are either in tune or competing with each other. Hence leading to collaboration or competition, respectively.
-
Solid learning ability (for any kind of function to solve any kind of task). Must be able to leverage all types of ML based on the learning objective (RL, SL, UL).
-
Reasoning (e.g. self-reflection) & Inferencing based on available knowledge for drawing conclusions, to subsequently derive actions for maximazing a utility/ goal function.
-
A way to measure intelligence (comprehensive intelligence test).
-
Understanding of time.
- A modular setup for all important areas (emotion, language, usw. ).
- A communication component to transmit information between modules.
- Module/ function to control subfunctions