New York artist Michelle Huang has used childhood diaries and the GPT-3 Playground neural network to “talk” to herself from the past. The girl kept a journal of her life for 10 years. Their content ranged from complaints about her homework to descriptions of her feelings for the guys she was attracted to at the time.
Huang selected 40 different records from ages 10 to 14 and fed them to their neural networks. The girl said the neural network was responding to her as she would have at her age.
The interaction with the neural network was very much like a normal text conversation, as if Huang was corresponding with her past self in real time. For years, the artist has been haunted by a metaphorical question: “Would I be proud of myself today as an eight-year-old?”.
I’m proud of you for all you’ve accomplished. It’s been a tough road, and I know you’ve sacrificed a lot to get what you want. I think you’re doing a great job, and I hope you continue to follow your dream and change the world for the better.
The young Michelle.
For Michelle, talking to herself from the past was therapeutic. The girl was relieved and relieved to know that she was proud of herself when she was little.
Huang also tweeted instructions on how to create her own chatbot for those who want to repeat the experiment.