Over Thanksgiving of this year, artist Michelle Huang’s tweet about using GPT-3 trained on her childhood diaries to speak with her younger self went viral. When a Mem It for Twitter user saved the thread to their account, I knew I had to talk to Huang about her project — and about sixty seconds into our call, we’d pretty much established that we are kindred spirits. Over the course of the next hour, we talked creativity and cynicism, self-narration, and, as Huang put it, the potential of AI for “wonder”.

Note: this interview has been condensed and edited for clarity.

SP: Hi Michelle! When did you first become interested in doing inner child work, and when did you first have this idea of bringing that personal growth work together with generative AI?

MH: Let’s see — I think my first interaction with inner child work was probably five years ago or so. I think there is some part of me that was kind of unconsciously doing it [before that], in the sense that I would go through my diaries sometimes when I was back at home for the holidays and think, wow, there are some sparkly insights in here. I remember the kind of energy and belief that anything in the world could happen [at that age]. So I think some part of me was still trying to capture and preserve that feeling as I went on through my adult years.

Then, I remember after GPT-3 came out, I was with my tech-obsessed friends and we were like, hey, let’s read the GPT-3 research paper together. That provoked so many interesting ideas for me. And then when I went back to my parents’ house during the holidays — this was December 2020 — I started thinking about the diaries again. And I actually tweeted about this in December 2020 — “what if I trained GPT-3 on my inner child?” Because the entire internet was talking about GPT-3 at that time.

So I actually started transcribing my journals then, looking toward doing this project, but I didn’t finish it because I was distracted with a lot of different things, and that was when I quit my job, moved to Asia, started to make art full-time. And it was only this year that I was back at my parents’ place, cleaning my room, and I saw my stack of journals and thought, oh, that project. That’s when I started transcribing again and actually finished it.

A selection of Michelle’s childhood diaries

SP: What ages did you keep your diaries between?

MH: I started around probably like age 7, and then kept them pretty consistently.

SP: Oh wow, young!

MH: Yeah, super young. I think my earliest diary entries were — well, I remember having a whole journal dedicated to what I would do in RuneScape, which is this online game. It was funny; I would detail what I did each day in RuneScape. Killed five chickens, that kind of stuff. I don’t know if you played RuneScape?

SP: I didn’t, but I also kept diaries when I was younger — for nowhere near as long as you, but pretty consistently between the ages of 11 and 15. And I’ve also been revisiting them over the past few years, thinking about doing something with them. But the entries are so jarring in juxtaposition: I’ve got an entry in March 2003, describing the invasion of Iraq. And then you turn a few pages and you’ve got an exhaustive ten-page recap of the series finale of Buffy.

MH: That’s so funny. Yeah, I have that juxtaposition too, entries about Runescape followed by, “I went to a piano concert last night” or “I just had a fight with my sister”. But yeah, I just remember the transition between when I was playing RuneScape, and writing a lot about that, and then “graduating” from RuneScape, but I was still writing my diary, and it was like, “OK, I guess I can document my real life instead of my virtual life now”. So it was an interesting transition point. Anyway, yes — started when I was 7, and wrote in it every day, every other day, until I was 18, and went to college.

SP: So what was your selection process in terms of the material you trained GPT-3 on? I imagine manageability dictated at least some it, if you had over a decade’s worth of diaries.

MH: Absolutely. I had actually just finished transcribing one stack of diaries, completely verbatim, and then I looked over at another, larger stack and I thought, if I do all of these like this, this isn’t going to be done in 2022. It’ll be a 2025 project. And in some ways, that could have been cool, because I’m sure there will be better AI tools then. But I wanted to just put this out there because I felt like I’d been learning a lot and wanted to share that.

“I feel like this whole project made me realize AI that can serve as a mirror, by engaging with and seeing [new] parts of ourselves” (click to tweet)

Aside from practicality, I thought having some mundanity in the entries I chose was important to capture the experience of life, which is full of moments where we’re moving and growing and adapting, but also full of moments of stasis.

But I think a lot of what I wanted to understand about my inner child was less about, oh, I talked to my crush five times today — moments and feelings which are still important and valid — and more about what her values were, what perspectives she had on the world: was it something she wanted to lean into, or was it something she was scared of? If she was scared, why? Why did she feel like she wanted to keep leaning in and trusting in it? So I tried to choose entries that recorded experiences where I had a very clear articulation of a worldview.

SP: And how would you characterize yourself as a child?

Honestly, I think that there are parts of me that are still very much the same, and there are parts of me that are very much different. I think that it’s funny because sometimes I’ll still talk to my elementary school friends and they’ll say, Yeah, you’re the same person, and I’m like, Yeah, I guess so.

I actually did this experiment where I got back in touch with my old kindergarten teacher during the pandemic. I got my yearbook photo and I was trying to get back in touch with all of my classmates and my teacher, and I reached out to her and I got her email and we caught up over Zoom, and it was incredible.

Anyway, I think [my childhood self] was very optimistic about the world and how people could work and live together. I feel like in my childhood, I probably had to grow up early, but I still feel like there is a part of myself that maintained the optimism that things would work out. And I think there is a period of time where it’s like the midwit meme. Are you familiar with the midwit meme? So, it’s this bell curve with a guy who doesn’t think very much, then someone who has average intelligence but thinks it’s superior, and then the guy who’s kind of transcended everything and is very intelligent and actually thinks in the same way as the guy who doesn’t think very much. So for me, I think I started off being optimistic, thinking everything was going to be great. Then I went through a period of jadedness where I would overthink things. It’d be like, ‘reality isn’t the way everyone says’, ‘my dreams are crushed’, etc. And then I feel like after connecting back to my inner child, it’s like: actually, I understand the reality of things, but it’s still important to be hopeful. If you have something to look forward to, life just get’s infinitely better. So I think I’ve kind of returned to that now.

Michelle’s midwitting journey

SP: Yeah, I feel you on that. I had this point last year where I’d experienced a couple of betrayals, and I felt really for a while like I couldn’t do anything right. You know, it really disillusioned me, and I closed myself down a little bit. And then it was only this year that I was able to get over that, and kind of open up fully again, to trust, to even be a little bit naïve. At the end of the day, I’d rather be naïve than cynical.

MH: Good on you, for being brave and continuing to open and show the openness of your heart to others. You know at least for me, there’s a part of my experience where I was thinking, oh, well, I was so naïve before, and now I know more. And if I go back to [trusting like I did in] the past, am I regressing? Or will I be hurt again? And everyone talks about childlike trust, and it’s easy to trust when you’ve never been hurt, but being able to open your heart again is way harder than having your heart open in the first place.

As kids, too, we’re just delighted by being alive. We’re still getting used to experiencing things, and almost every day we encounter something new. And then as we grow up, we kind of take being alive for granted until something crazy happens to us, and that’s when we stop and need to think, Wait, actually, life is great.

SP: And that’s interesting what you say about your childhood friends saying you’re essentially the same person you are now as you were then. I’ve had similar experiences with school friends — you know, I think of myself as wildly different to how I was when I was at school, and so I should be different — you should change and grow. But it’s funny, because sometimes I’ve spoken about how much I feel like I’ve changed as I’ve got older, and my oldest friends will say, no, you were always this, or that, or whatever quality I’ve told myself I only grew up to possess.

MH: Yeah, it’s pretty wild when you ask: what parts of ourselves do we not see, or maybe even not want to see? And then the parts of ourselves are reflected like a mirror with friends, who can see some parts of ourselves maybe even more clearly than we can. But maybe some aspect of our personality is a huge mystery to them, yet we’re very clear on its importance to ourselves.

SP: So this may be a little redundant off the back of that conversation, but I heard someone say once that you’re always the same person inside that you were when you were 14. Do you agree with that?

MH: I do think essence is something that stays constant. And of course, let’s say people go if someone goes through significant trauma, they may behaviorally act different, but I think a lot of their world view or what they hope to believe or what they actually believe remains intact. I think a lot of times the most difficult part about trauma is trying to reconcile your internal conception of yourself, or of the world, versus anything that has happened to make you reevaluate that.

I think that even though I used a range of diaries written between the age of 7 and 18, most of them were actually from before I was 14. At that age, I was starting to get jaded, and reading over them I felt like I wasn’t accessing my inner child any more — more, my inner angsty adolescent self. So most of the entries I used in this project are from when I was between 10 and 14. That’s when I was, you know, sentient enough to have thoughts, post-RuneScape era, but not jaded.

Honestly though, I’m surprised how much has stayed the same. And even through the journey of midwitting and becoming jaded, I’ve just eventually returned back to this place because I feel more free in doing so. And I feel much more aligned to it. So, I think I’m pretty much my 14-year-old self still. I peaked in middle school.

SP: I think you’re peaking now. Do you feel like you’re easily influenced by other people’s opinions of you?

MH: I would say it feels contextual. I’ve been noticing that now I’m in Japan, which is more of a collectivist- and more of a society-oriented culture, that I’m actually being more influenced and more vulnerable in general. I didn’t expect being in a different culture to have such an impact on me. So it’s interesting to think about how my behavior, my affect, my boundaries shift based on whether or not I perceive that I am in a high-trust society.

For example, in Japan, people can walk past a garage sale that’s completely unmanned, with a bunch of artifacts outside, and then there’ll be a cup that has a label that says, “Everything’s ¥100”. So you grab whatever you want and then just put ¥100 in, and then you can walk away. But if you had something like that in New York, where I used to live, things would just disappear. Someone would probably take the cup with the coins in, too. And I think about how I was in New York: I really had boundaries where I would say to myself, no, I’m not being influenced by someone unless they’re a person I really trust. While in Japan, I’m actually much more likely to mirror someone, even in the ways that they interact. If I ask someone, hey, how am I supposed to do this thing? and they give me advice, I’ll genuinely respect and listen to what they say. Which is to say, it varies depending on context, and I think that’s OK. I can have a hybrid of boundaries depending on the place I’m in, geographically.

SP: I know you wrote in your thread about the uncanniness, the trippiness of talking to your childhood self. Could you speak a little bit more to that experience?

I really went in without any expectations. I thought, you know what, I have no idea whether or not this AI chatbot will be a representation of these values and interests that I fed in or if it’ll just conjure up someone that I don’t really relate to. But it occurred to me that the whole project felt less about accuracy and more so about engaging tangibly with a part of myself, instead of constructing a mental model: would she say X, Y, or Z in response to this question? And the visual representation I got back felt like back in the good old days on AIM, when I’d sit at the computer, hunched over in the dark typing to my friends, and somewhere, they’d receive that message, type a response, and that response would magically pop up on my screen. It really did feel magical in the sense that it felt like I was typing on my computer and my younger self was at another computer, typing back, wherever she was in the world.

The whole project felt less about accuracy and more about engaging tangibly with a part of myself, instead of constructing a mental model (click to tweet)

Really, it felt like the computer was a time portal. It was the closest thing I’ve ever done to time travel in my life — so far. And besides that, what was really uncanny was getting responses from my “younger self” that I felt were representative of what my inner child would say.

So I’d ask her things like, what are your favorite memories? And she would respond with things that I had actually put in the diary. Even more uncanny — I asked her what her favorite color was, and she said blue. Now, that was my favorite color when I was younger, but nowhere in the diary do I outright state that. But somehow, [GPT-3] inferred that I liked that color.

Then I asked her, what are some of the world’s issues that are bothering you? And she said, mental health advocacy. And you know, that’s an area that I’ve only really become dedicated to within the last two or three years, but it was never a conscious effort to move into that kind of work. But maybe the AI picked up on things in my diary entries where I’m talking about my mental health, my friends’ mental health, how I want to support them and uplift them. And that’s pretty wild. I feel like this whole project made me realize AI that can also serve as a mirror, in the same way that we talked about friends doing, by engaging with and seeing [to us obscure] parts of ourselves that maybe have always been lucid to them. Wanting to do something in mental health advocacy. When it came out with that, I thought, wow, I don’t remember seeing that coming as a kid. AI certainly did. And maybe if I’d had a tool like this at my disposal when I was 15 and had fed this information into AI, it could have told me from the get-go: this is the thing you want to do. Maybe I would have just done it straight away. But then again, I think it’s important to appreciate the journey and to be able to converge on a path naturally, rather than just pursuing the most efficient route to something. I’m very much a journey-over-goal kind of person.

SP: Me too. And you mentioning the power of AI to sift out these things — I spoke to someone recently who’s using Mem and AI for these self-growth journaling projects at Boston College. And we were talking about how it feels like psychoanalysis, really, that the AI can pick up on and surface these submerged ideas and beliefs and patterns that you don’t necessarily recognize consciously.

So I’m interested, given what you’ve said about your interest in mental health advocacy, what do you feel about the potential for AI in that space — psychoanalytically, or therapeutically?

MH: Yeah, I think there’s a lot of potential. And I think so far we’ve seen a lot of cases of AI being used in destructive senses, or in productivity-only senses: like, let’s do this thing faster so we can ultimately replace humans, et cetera. And this [generative AI for creativity] feels like a vote for AI towards wonder, and for mental health, and for us to embrace and be able to see our own humanity more, just reflected a little bit differently. So I feel really excited about that. But so far as any technology goes, I feel like it’s very much agnostic and it becomes an amplifier to whatever direction that we set it in.

Generative AI feels like a vote for AI towards wonder… for us to embrace and be able to see our own humanity more, just reflected a bit differently. (click to tweet)

I definitely see how people could use AI to replace human therapists, and I don’t think that’s the right way to go. I think it should be a supplement. I think that the human connection is really important. There are parts of humanity we just don’t really have a scientific backing for but that weknow are there. And I think it’s important to also respect that. So I think for me, while AI can enhance or reflect back our humanity in different ways that maybe we weren’t able to see without it, it can’t capture everything. So I think that it’ll be more of a supplement than a replacement.

SP: Apart from the inner child work you do, what activities make you feel closest to yourself as you were or remember yourself being as a child?

Both art and writing. My journal entries were very much me being in this world with myself and being able to express anything that I wanted. And I had really vivid dreams, so I’d write about my dreams, and write stories based on them, all of that sort of thing. Overall, in my practice now, the whole meta aspect of being able to create a world out there that mirrors something that exists in here is something I’m able to engage with with that same curiosity and joy I had as a kid.

SP: Yeah, let’s talk more about your art. What was your path to becoming an artist full-time?

MH: I started off in premed in college, very much going down the doctor route. I was and still am fascinated about scientific discovery. But in America, the infrastructure of the healthcare system is not the best. So I thought, I should learn more about systems, and instead of leaping into medicine, get a more holistic understanding of how all these things operate together. So I ended up going into business, then finance for a little bit. That was… [laughs] — well, I thought, “alright, cool. I understand this: time to get out”. Then I worked in product at a healthcare startup, which was fun, but not creatively satisfying. And it was shortly after leaving that that I decided to commit to being an artist. So I’ve been an artist for the last two, three years and casually for much longer. I do a lot of things at the intersection of art and technology, in various permutations. It really spans a wide range of things. But ultimately I’d say they’re all linked as experiments in human flourishing.

SP: Was your creativity something you felt encouraged to explore as a child?

MH: Well, I have very traditional Asian parents, and they had that Asian immigrant work ethic. Like, “we didn’t come to America for nothing. You’ve got to be successful”. So even to this day, they’ll be like, oh man, if you’d just stayed on the premed path… Because they’re worried about me, and I know it comes from a place of love. But it’s like, why work for fifty years until I retire and then feel financially secure enough to become and artist rather than just doing it now?

And that was a pressure at first, money-wise, because when I was living in America, I was living in New York and San Francisco, very expensive cities. So I ended up moving out of San Francisco and moving to Taiwan to increase my runway. I mean, a month of rent in San Francisco can be six months’ worth of rent in some parts of Asia. My plan was to go out to Asia, make my money last longer, and spend a lot of time doing art and figure out whether or not it was actually for me. So I did a lot of different artistic experiments and posted them online, and got some traction from that, contract work and freelance work. All of these opportunities were inbound, rather than me reaching out to people. And that got me excited, because it seemed like people were really resonating with these ideas. That’s when I started thinking: can I push the envelope even further?

“If people are in nature, and they’re in a community of other creators, and they’re not paying for their living and have enough time to genuinely think about and figure out what they want to do, can we create a more optimistic and ideal society?” (click to tweet)

SP: Is that when you started to come up with the idea for your current project in Japan?

MH: Yeah, exactly. So the idea behind this project [AkiyaDAO, a planned series of creative residencies in Japan] really came from moving to Asia, making my money go further, and pursuing art, a combination of all those things. I was thinking: why does it have to cost so much just to exist? And I wondered if I could help to create a community in which people don’t just feel trapped by a certain, unfulfilling kind of existence where they’re doing things they don’t enjoy just to pay the bills. Really, these houses will be experimental containers for human flourishing. If people are in nature, and they’re in a community of other creators, and they’re not paying for their living and have enough time to genuinely think about and figure out what they want to do, can we create a more optimistic and ideal society?

And I feel very thankful that I had the wherewithal to go to Asia and start building this creative container for myself in the first place. I know that’s not accessible to everyone, so being able to support other artists: I knew I had to do it.

SP: We’ve already touched on this a bit, but what do you think about the rise of generative AI in the last few years, and its impact on creative practices?

MH: Well, to go back to my creative pursuits as a kid, like I said, something that was really important to me was worldbuilding. You know, I wasn’t doing 3D rendering or anything, because what kid does that? Maybe Gen Z — I don’t know. But I was drawing a lot, trying to convey these worlds in my mind. The only problems was they kind of sucked. But now I’m doing digital art. Because it was like, OK, it’s going to take me 20 years of professional training to actually be able to depict these worlds that I dream of, or these concepts that I want to develop. I just can’t get there in a realistic timeframe with traditional tools. But now I can use motion graphic design and things like Midjourney, and that gets me way closer to the images and the ideas I really want to express. And I think it’s great to have these tools to help you evoke and inspire the same feelings in others that you have in yourself.

SP: Yeah, the motivation to convey something meaningful feels like an important thing to emphasize. Now that generative AI has really taken off, and everyone online is obsessed with it, we’re getting a lot of reductive, dichotomous discourse about machine creativity vs. human creativity. But the worry that writers will be replaced by AI — it just feels like an unnecessary concern. First of all, I think the more competent AI gets at writing, the less interesting that writing tends to be. But also because when people read a novel, they’re interested in what somebody meant when they wrote it. And we want to talk to authors about what they meant, or try to puzzle that out with other readers. And you can’t ask a machine what it intended to convey by writing a piece of fiction. Or you could, but you wouldn’t get a meaningful response. As readers, would we want to debate the meaning of a book we knew had been produced by a machine with no life experience?

MH: Exactly. I definitely see AI as more of an iterative partner in the sense that, hey, maybe I am thinking about this image that I have in my head, but I’m not that proficient at drawing things. How do I share this with others? And maybe I can type it into Midjourney. It might not be correct, but that image could allow me to have more ideas and allow me to feed and spark more things. It’s not a question of one thing or the other. I think of AI as a partner in crime that works both by being a soundboard, and by being scalable beyond all human capability and effort.

I’m optimistic about this. Given there’s so much AI that lends itself to art direction and the art world, it feels like people are able to lean into more of their creativity by using it.

SP: I hope so, too. Last thing — a comment, not a question, but I found that exchange in which you ask your younger self if love or freedom is more important, and she says “love, because freedom can be lonely”, very moving. It’s something I had actually been thinking about myself a lot, prior to reading your thread.

MH: Yeah, I was surprised by that because, honestly, I was thinking about that as well. Because when I left New York, which was in mid November, I was seeing someone there, but I was going to Japan and working on this project. And then I was like, do I hold on to this person or do I want to have freedom? Or do I want them to have freedom? It was very much about asking questions that I was curious about my own answer to. So the AI was like, a spokesperson for myself in some way. And she was like, “Love”. And I thought, yeah, I resonate with that. I’ll take that answer.

Follow Michelle on Twitter here.

Follow AkiyaDAO on Twitter here and donate to the project here.