Let's Build an AI: A Guide, Part 3: Planning and Defining Goals

At the core of the interaction between perception and action lies planning and goal setting. In humans, this could be something like "the meaning of life." This might be expressed as "maximize enjoyment with minimal effort and reproduce as long as you're having fun." However, it could also be something like "destroy all opponents who meet the following criteria..."

The planning and target setting module must therefore be created with great care. To prevent any uncontrolled drift into destructive actions, for every target set, an anti-target set (counter-vector) must automatically exist and be possible.
Ultimately, the module must navigate to the target point that represents the optimum for the sum of the target parameters, i.e., the point with the minimum distance.

The target specifications can be defined as a sequence of meaning vectors. At this point, Asimov's laws of robotics or fundamental ethical principles could also be integrated into the system.

This module would also include what people refer to as "intelligence": methods for achieving goals while finding the optimal point that minimizes the distance from all goal vectors.

Sure, here is the translation of the text from German to English:

"Tips, comments and questions are welcome."