How to solve the problem of Artificial General Intelligence (AGI). Part - XI : The General Theory of Intelligence

The General Theory of Intelligence is a lot like the General Theory of Relativity. It's main goal is to solve the problems of narrow intelligence models. The field of Artificial Narrow Intelligence (ANI) is littered with many ad-hoc procedures and rule-of-thumbs to get a particular job done. This line of thinking is very different than what the earlier scientists of AI had in mind. Artificial Intelligence as a science is very much concerned about finding a general scientific principle which will facilitate creation of human-like artificial entities.

The theory uses the universal priors derived from physical laws. The immutability of laws of physics and the physical constants plays a major part in this. The limit of human intelligence and also the limit of computing in the same sense depends upon ultimately on the physical constraints posed by nature.

Any physical understanding is possible through the clues provided by nature. For this reason the universal priors reflects the realities of the world for constructing proper world models. The world models are both external and internal at the same time and works on feedback mechanism between them.

In this theory consciousness may be modeled as a meta-level like the hyperparameters of deep neural networks or some other parametric models. Human-like thoughts could be generated merely as the reflection of external reality in this model which corresponds to a higher level control.