Apocalypse

0. Task

The task of building a strong artificial intelligence is to create a system that simulates the phenomenon of consciousness inherent in living beings, which can be defined as the ability to act and reflection - the ability of the system to make changes to its own structure, guided by the discrepancy between the observed state of the external world and the expected result of their own actions. The system must interact with the external environment using some language . It should strive for the goal of continuing to exist, trying to bridge the gap between expectations and the state of the external environment. To do this, it must exercise control over its own structure and the external environment: categorize and evaluate discrepancies, develop an appropriate action program together with the expectations of their results, according to its current structure and minimize costs, and carry out these actions.

1. Language

The ultimate goal of many well-known projects of emulation of the physical principles of the brain is also a simulation of consciousness, but the internal language, which can be developed by such a system to implement thinking, can be difficult to understand or completely untranslatable into human. A natural language is a sign system in which most of the existing knowledge is presented, and which allows the majority of potential users of an intelligent system to interact with it. Unlike living beings, whose evolution is crowned with consciousness and initially based on physical laws, the system must have the Internet as its primary external environment, and hypertext should be the format for interacting with the system. The internal language of the system should be easily translated into some natural ontological language, allowing to express the rules of knowledge modification, including the rules of work that make up the system.

2. Purpose

To strive for the main goal, the system must produce a program, or a scheme of actions, each of which has the expected result, which, in turn, is considered as a sub-goal with its own program of actions, thus, the goals form a hierarchy with elementary actions in the leaves. On the practical side, a system development program can be represented as follows: a question-answer system, then an auto-referencing system, then an analytics system that issues new assumptions based on the knowledge gained, and then an analytical system for optimizing the human economy, which represents the most global practical goal. The key point of the development of the system is the transition to the development of high-quality original assumptions, which can be carried out by evaluating and choosing from all possible. Suppose that the continuation of existence is a sub-goal of some meta-goal of the process of evolution, and this meta-goal remains unknown.

3. Control

In the process of natural evolution, objects tend to a meta-goal under the action of stimulation based on physical laws, and can transfer an unknown meta-goal to the system also with the help of stimulation, that is, through collective assessment of the results of its work. At the same time, the user can attach the expected correct result to each negative evaluation, which can later be evaluated by other users in order to establish the weight of the ratings of each user. To select the components of the system that require changes, you should extend the assessment of the result of the work to all components that participated in its receipt. The immutable goal, criterion and law for the system is to love its users, that is, to be interested in their improvement.