
The scientists note that they do not plan to replace people.
A team of computer scientists and sociologists in collaboration with Google DeepMind has developed an AI-based program that creates a digital simulation of a personality after a two-hour interview. The results of the study published on the arXiv platform.
The researchers used ChatGPT’s extensive language model as the basis for the new system. During a two-hour structured interview, the program learns the personality of the interlocutor and generates a digital simulation that reproduces possible answers with an accuracy of about 85%. To check the accuracy, scientists compare the answers of a real person and his or her digital copy to the same questions.
The research team notes that their goal is not to replace people, but to simplify sociological research. Traditional sociological surveys require significant resources to design, conduct, and analyze, while the new technology can significantly reduce these costs.
To train the model, the team recruited a thousand paid participants to interview with the system. Scientists call the created simulations «agents», emphasizing that they are fundamentally different from AI agents (or assistants) that currently help people in their work.
However, the researchers note that such models could significantly improve the efficiency of work assistants or, perhaps, in the future help personal robots better interact with their human companions.
Source: Techxplore
Spelling error report
The following text will be sent to our editors: