Imagine an AI model that reproduces your behavior and personality traits after a two-hour interview. According to a recent study by Stanford University and Google DeepMind specialists, this futuristic concept has become a reality.
As LiveScience reports, the researchers created over a thousand copies of real people. The participants in the experiment were asked questions about their life priorities, values, and views on social problems. In addition, the participants performed logical tasks, which allowed the researchers to gain a deeper understanding of decision-making mechanisms.
This information became the source material for the creation of “modeling agents”, essentially AI replicas trained to reproduce human behavior with impressive accuracy. The AI copies reportedly matched the responses of real people by 85%, especially in personality questionnaires and predictions of social attitudes.
However, the digital copies of people struggled with interactive decision-making tasks, where the AI had to make choices depending on the situation, copying human reactions with less accuracy.
The authors of the study, summarizing their work, highlighted the potential benefits of using such technology. These digital clones could be used for social experiments, predicting reactions to social events, or studying the impact of social institutions. The technology will also likely be used in the future to copy the personalities of relatives who may soon die.
But, of course, like any breakthrough technology, AI copies raise serious concerns. The ability to duplicate someone's identity based on just 2 hours of data could lead to potential privacy breaches, identity theft, and “technological” exploitation. Therefore, scientists emphasize the need for strict controls over the development and use of such technologies.