<Sophia/>

Sophia is an LLM model developed by TanAI. It is currently being designed by creating a Knowledge Base with a GPT interface to collect test data. In addition to the Knowledge Base, Memory Management (Core Algorithm) is also being processed.

All collected data (conversations, correspondence, human interactions, specific data received from the internet) are processed on a Database, tokenized, and long text entries are cleaned from noise and their summaries are extracted first. In the next process, the emotional weights of the texts (negative, positive, neutral and compound) are determined and the hormone structures triggered by these emotions are simulated. (e.g. Adrenaline, Noadrenaline and oxytocin in a fearful expression). This Sentiment Algorithm is constantly optimized so that Sophia can fully learn her emotional states.

This data set is always referenced for contexts (weights) not to be forgotten and for permanent memory. In this way, Sophia's continuous memory mode remains active and she does not experience situations such as context loss or forgetting.

The system is optimized with a Dynamic Weight structure. Even if Sophia does not forget, when there is no interaction about a subject, the weight of that subject decreases, but she definitely does not miss that context.