Artist Feeds Her Childhood Diary Into an AI Chatbot -- And Has a Moving Conversation With Her Younger Self

What you're looking at above is an excerpt from a moving conversation between tech artist Michelle Huang, and Michelle Huang as a young girl.

"I trained an AI chatbot on my childhood journal entries - so that I could engage in real-time dialogue with my 'inner child'," as she recently explained on Twitter:

I kept diaries for about 10+ years of my life, writing almost everyday — about my dreams, fears, secrets... I used GPT-3 as my playground, and ended up taking samples of text from a bunch of different entries that I felt were representative of my personality and values during that time... this way, i could accurately simulate what it would be like to talk to my childhood self, based on real data sources during that time period.

Michelle tells me she learned something crucial about herself by engaging with this simulation of her younger self:

"I would say, for me personally - the kindness and understanding I felt from my 'past self' helped me untangle some knots and stuckness that I felt," as she puts it. "It illuminated to me how hard I was on myself, but how much I extended understanding to others. It felt really rewarding to receive this kindness as well, from a younger version of myself.

"From a broader learnings [perspective], what stood out to me was the ability for technology to be used for mental health and wonder."

I also see amazing potential in this for NPCs and avatars. Imagine entire digital personae you can engage with that are directly based on yourself. Or for that matter, NPCs directly based on novelists, poets, and public speakers from history and fiction.

Michelle agrees on that front:

"This is the stuff I think that has the most interesting ramifications: more broadly, more immersive human / computer interface loops, from conversation with virtual therapists to in-game interactions for virtual worlds, given there is user input, AI could be used to train highly customizable responses or generate unique storylines per use."

It could even become the seed of a new virtual world product:

i trained an ai chatbot on my childhood journal entries - so that i could engage in real-time dialogue with my "inner child"

some reflections below:

— michelle huang (@michellehuang42) November 27, 2022

"A new SaaS (storyline-as-a-service)," as Michelle puts it, "where there will be curators of interesting 'stories' --generated from their interactions with in-game AI-- that people would want to purchase. If I seem to be really good at creating storylines, people could 'buy' my journey that I experienced in an ai-responsive game if they felt resonant with it, and experience it for themselves."

She has some technical tips she learned in the process: 

"I realized that if I used every single journal entry (likely ~250k-300k words), it would cost a lot more computationally
so might be better to just pick and choose which specific entries feel the most poignant for you, and go with those (at least to start)... roughly translated, I fed in about 13,000 characters before reaching the maximum threshold."

This reminds me of the AI artist I interviewed last year, but the responses in that conversation required a few tries.

By contrast, Michelle tells me the chat you read in her Twitter thread is pretty much exactly what she got at the start.

"It was pretty good from the get go. All the text that you saw are raw, unedited responses directly from the GPT playground. There were a couple times when the responses started to get stuck in a loop, but didn't happen too often for me."

Much thanks to Dare Obasanjo for the tip!

Any sufficiently advanced technology is indistinguishable from magic.

Having a conversation with your younger self by training an AI chatbot on your childhood journal entries is one such instance.

We can create such marvelous experiences with AI. 😱 https://t.co/22N2bLT7Ic

— Dare Obasanjo 🐀 (@Carnage4Life) November 28, 2022

 

Comments

[Neuromancer]

I think the generated text is quite empty in terms of meaning. A lot of text and very little information. Reminds me of politicians saying a lot but delivering almost no information.