GPT teaching assistant as a tool for improving teacher-student interaction

Abstract

Neural network systems have recently been used for various purposes. However, direct application in the educational process, as practice shows, is difficult. The authors propose and implement a new ideology for the use of neural networks in teaching. Using chunking technologies in the neural network-lecturer dialogue ensures the creation of a quasi-model of a real teacher. This approach provides significant savings in tokens, optimization of the knowledge base volume by limiting only relevant chunks, and limiting the time consumption of communication with the system.