Human Personality Can Be Replicated with a 2-Hour Interview with AI
Date
Nov 24, 2024
Classification
- AI/IT/Productivity
#
- Generative AI
(Photo = Shutterstock)
Shuka Left Too? Shaking MCNs and Creators Choosing 'Independence'
🫑 3-Line Summary
•
Researchers from Stanford University and Google DeepMind announced 'AI digital twin' technology that mimics an individual's personality and judgment with an 85% match rate using only two hours of interviews.
•
Unlike conventional methods of collecting simple demographic data, we have demonstrated that in-depth qualitative interviews can precisely replicate an individual's unique experiences and ways of thinking.
•
This will be an innovative tool that will change the paradigm of future public opinion polls and social science research, but at the same time, we must be wary of ethical risks that go beyond deepfakes.
🥦 Insight
•
Are you exhausted from all-day meetings and decision-making? According to this research, you might soon be able to create an "AI twin" that looks exactly like you and send it to meetings in your place. It would essentially mean having an avatar that says, "If it were me, I would have decided this way," with 85% accuracy.
•
However, here we must ask a spine-chilling question: "If 85% of my judgment can be replicated with just two hours of data, where does my irreplaceable value lie?"
•
Technology will take over the repetitive decision-making we dislike. In that case, entrepreneurs must focus on the remaining '15%' that AI cannot replicate. This includes things like intuition that cannot be explained by data, sincerity that moves hearts, and the courage to make irrational yet great decisions. Isn't it time to spend time building an 'unreplicable self' rather than 'technology that replicates me'?
— View original —
A study has found that a two-hour interview is sufficient to understand an individual's memories, personality, and tendencies using artificial intelligence (AI). This suggests that it is useful for building virtual replicas of humans, or digital twins.
A research team consisting of Stanford University, Google DeepMind, Northwestern University, and the University of Washington recently published a paper titled "Simulation of 1,000 Generative Agents" containing this information on RAIX.
The research team recruited 1,000 people of different ages, genders, races, regions, education levels, and political ideologies. They paid up to $100 for participation.
The research team created agents that were replicas of individuals through interviews. To test how well the agents mimicked humans, participants took personality tests, social surveys, and logic games twice each at two-week intervals.
And the agent completed the same exercise. The result was 85% similar to a real human.
Park Jun-sung, a doctoral student at Stanford University who led the research, said, "Little 'yous' can go around and make decisions that you would actually have made," adding, "I think this is ultimately the future."
Simulation agents make it easier for researchers in the social sciences and other fields to conduct studies that would be costly, impractical, or unethical to use actual human subjects. These are also referred to as human digital twins. Companies like Brox AI and Unlearn are already building human digital twins and utilizing them in polls and clinical trials.
Want to read the full article? Click!
Please click (touch) the link block below.
Go to the webpage where the article was written.