Training and functioning of AI

🚧 Content is currently under revision 🚧

Do I have to train the AI manually?

Every conversation conducted in live operation adds to the AI's self-assurance. For example, if a query is answered positively in the case of uncertainty, this is fed back as a positive signal from the AI. These inputs are processed in the reassuring phase. Thus, each reassuring provides a changed and optimised understanding of the AI. If the user denies, gives negative feedback at the end of the conversation, or repeatedly makes rephrased requests, these signals are also fed back into the AI into Dreaming. These are used by the AI in two ways.

  1. to increase confidence in choosing the appropriate response.
  2. to highlight new request topics and suggest new answers

This means that no manual retraining is necessary per request.

"Learning by doing”

Can I still train the AI?

For faster learning and understanding of the AI, further variants can be added to the AI at any time. This expert input is taken into account in the dreaming phase.

The AI can be manually trained by experts for faster and more accurate results. 

"Learning from experts”