AI has learned to clone a personality in just two hours – and this is a dream come true for sociologists and deepfake scammers.
Scientists from Stanford University and Google DeepMind have collaborated to develop an AI application capable of creating a digital replica of a person’s personality.
Their findings, detailed on the preprint server arXiv and reported by MIT Technology Review, show how generative AI can mimic human traits with remarkable accuracy.
The system, built on the large ChatGPT language model, works by asking a series of questions, collecting answers over a two-hour session, and processing them to generate a digital copy of the person’s personality.
To test its accuracy, researchers asked identical questions to the study participants and their digital “twins.” The AI-generated personalities gave answers that matched those of their human counterparts 85% of the time.
“If you can get a bunch of little ‘yous’ running around and actually make decisions that you would make, I think that’s ultimately the future,” said Joon Sung Park, a computer science graduate student at Stanford, involved in the research.
The developers emphasized that their goal is not to replace people, but to streamline tasks for sociologists and researchers. They noted that conducting surveys of living respondents can be both time-consuming and expensive.
This groundbreaking work underlines the multifaceted nature of human personality, which seems immeasurable and complex.
However, the researchers have shown that an individual’s worldview and value system can be effectively inferred using a well-designed questionnaire. Once this data is fed into generative AI, it can convincingly imitate personality traits – at least within controlled environments.
While this innovation has exciting applications in fields such as sociology, it also raises concerns. If AI can convincingly replicate personalities, it could make deepfakes and other forms of digital impersonation more sophisticated and potentially dangerous.
