Michelle Huang, an artist-scientist from New York has found a way to train the AI chatbot ChatGPT to engage in real-time dialogue with her younger self.
Huang fed the AI behind the chatbot, which was OpenAI’s GPT-3, with her childhood journals, and got responses”eerily similar” to how she thought she would have responded then.
By asking the chatbot questions and letting the chatbot ask herself questions, Huang has gained a better understanding of her younger self and, in turn, got confirmed by her past self.
How did she feel after…
Huang shared her feelings after chatting with her younger self. It was deeply healing and therapeutic.
these interactions really elucidated the healing potential of this medium: of being able to send love back into the past, as well as receive love back from a younger self the stuckness becoming unstuck, of finding closure with past guilt or stories that we had of ourselvesMichelle Huang
Many are exhilarated by Huang’s findings and are excited about ChatGPT’S potential in the field of mental wellness therapies.
How did she do it…
Many people asked Huang to share how she managed to make this happen. So she shared how she “trained” the AI in a later tweet.
The process could be summarized as acquiring source material that reflects your “personality/voice/values” first, like writings, diaries, or chat history. Then feed the materials to the AI through GPT-3 playground and set your preferred training models.
The next is to craft a prompt in the playground and play with the parameter settings to set a tone. Once you hit submit, you can talk to your trained chatbot. And you could always refer to Huang’s original post for the step-by-step tutorial.
In case you get confused…
You might get confused with ChatGPT, the chatbot, and GPT-3, the AI behind the bot. The following interview with AI expert Chris Rowen might help you to understand the topic further.
Always ruminate and get stuck in the endless loops you are desperate to break through? Why not give AI a try if you have a dusty pile of old diaries or, even more conveniently, some archived chat history?