ai tech

Disassembling a character

A character, whether created within a movie or a book or is indeed a human, is based on multiple layers of grouped core elements. To reconstruct an existing character, we initially deconstruct the elements of our target character and reconstruct them, adding informational elements from multiple sources. Information itself does not create a character; it only creates a Persona that knows a lot about a specific subject, real or fictive person. 

How a Persona is built

A Persona, or in other words, reconstruction of the character of someone or a figure, represents numerous layers built from grouped characteristics in various areas. 

 A specific combination of different elements builds a layer of a character. Further, each layer element has a variable intensity that is flexibly applicable based on the input, situation, interaction, necessity of producing results, and many other factors. 

Artificial character versus human

Artificial characters are very similar to characters of actual humans, with the difference being that they lack the complexity and number of character layers a human has. This is attributed to the fact that when humans observe or reconstruct an external character, they simplify both the elements of a layer and the number of layers that build the total character due to limited understanding, observability issues, and barriers to using precise language to describe the character to be analyzed. 

Humans contradict themselves

A further important element of the reconstruction of a character is the understanding that different layers of a character may produce contradicting positions, guidance, and insight based on external information or stimuli. This fact has to be embraced and not balanced out, as it represents a standard human condition of uncertainty about the right thing. Building a character as close as possible to the original involves a level of uncertainty, representing the reality of life, human and nonhuman decisions, and coincidence. 

Layering and dynamic processing

With our unique layering technique, we break down a character into its core essential elements, sort them into groups, and divide them into various categories through a dynamically applied process, continuously evolving our training model. We avoid large training units by training individual layers or smaller parts like the core elements.

This produces a much higher level of flexibility and fast retraining of parts of the AI. It can be regarded as continuous training and retraining as it happens on different layers.