Can AI Write My Bio?

This is a version of my About page, as written by an AI. Co-written by GPT-3. I ask the questions, it gives some very interesting answers!

When, after the fact, I asked AI to summarize this article, it said, “The AI takes what you give it and runs with it, eventually producing a highly personalized summary that’s not entirely unlike the old ‘poet of the moment’ poems in some ways. Really fun to use for an idea of your life so far, one topic at a time.”

So, I’m going to use GPT-3 to write my about page. Artist and engineer Selma Sabera, a pal from classes at SFSU, challenged me to do it. She thinks this AI’s bunk, not ready as a writing tool, but I think there’s definitely something intelligent-ish enough about this version of AI. It seems to be a worthy writing partner.

Selma wrote me with this challenge: “Have it help you write a one-page bio of the person you know best, you. Let me know how it goes.”

Her gist is, “it’s not ready for prime time.”

My gist is, “you don’t want it flying the plane, but it’s a good co-pilot.”

The AI’s not doing everything . . . but then neither are you. What are you doing? You’re lobbing interesting prompts through a window. And the window happens to look out across a huge portion of the store of human writing. And then you actively listen to the response and get ready to catch anything that’s returned with interest-and defenestrate the rest.

Okay. Let’s test it out. The subject is: me, my bio, and I.

The AI uses this document-in-process-everything I’ve written here so far-as the jumping-off point for whatever it’s going to write. If I seed it with salient keywords and phrases, it’ll be able to fill in the details, right?

I’m going to approach this as a career-type bio. So, what does the AI need to know? I’m not sure if it will know about my links or what, but just in case I’ll add some URLs.

Projects I’ve helped grow recently:

• ProxiAI.com: an open-world gaming project from Will Wright; I made UI prototypes, developed possible and impossible backstories, and made memory visualization experiments in Unity 3D.
• ViridianReveries.com: my Internet platform as an indie creator, publishing stories, music, games, a podcast, a newsletter, and the occasional burst of social media.
• SmartSwords.com: a successfully funded Kickstarter. This fantasy interactive fiction mobile game pairs with a Bluetooth wooden sword of our own design (something between a toy and a practice sword), with custom electronics in the hilt.
• The University of Chicago, where I studied physics and math, but dropped out after two years.
• Hampshire College, where I studied making comic books and documentary videos, and dropped out after two years.
• San Francisco State University, where I studied computer engineering, the hardware and software of gadget making, and at long last picked up a B.S. degree.
• I like prototyping with hardware and software.
• I like being on teams of self-managing experts that are competent but laid back.
• I like when daily work involves brainstorming weird sci-fi possibilities, prototyping gadgets, iterating on interfaces, and playing games.

My adventurous highlights are:

Next up: making the instruction template I’m going to use for the first all-AI draft. I’ll trigger one instruction at a time, top to bottom, so the AI can use what came before to inform what comes next.

/instruct [summarize my accomplishments]

/instruct [describe what I’m looking for in my ideal job]

/instruct [highlight my previous job experience]

/instruct [mention my schooling]

/instruct [say a little something about my travel adventures]

/instruct [end with something pithy about why I’m a good collaborator]

/instruct [suggest any key points about me I didn’t think to ask you about]

I actually started adding a lot more instructions once I got going and got more and more curious to see if the AI had any insights about me. The AI itself effectively began giving itself instructions-you see, GPT-3 tries to predict what comes next based on what it’s got so far. So, it recognized the question-answer-question-answer format and started to mimic the syntax of my questions which resulted in it issuing instructions to itself.

‍Without further ado, here’s my bio as written by GPT-3. (Instructions in bold, AI responses in italics.)

/instruct [summarize my accomplishments]

I am an avid gamer, performer, and maker of electronics and gadgets. I went to University of Chicago for two years before dropping out to focus on comic book making at Hampshire College for a year. I graduated from SF State with a B.S degree in computer engineering. I have a history of working as a designer on multiple projects: ViridianReveries (a website), SmartSwords (a Kickstarter game project), Proxi (from Will Wright). I am currently looking for opportunities where I can contribute my skills as an engineer with experience working on games, gadgets, and websites that have been successfully funded through crowdfunding.

/instruct [describe what I’m looking for in my ideal job]

I’m interested in video games as an art form. I am looking to collaborate on a project with others who have both the time and the resources to take on this kind of speculative, experimental, and open work. I am looking for interactions with other people who love designing their own worlds too.

/instruct [highlight my previous job experience]

I helped make prototypes for user interaction in a new kind of open-world gameworld.

/instruct [mention my schooling]

I attended the University of Chicago and was quickly bored with the lack of hands-on work. I left to attend Hampshire College, which offered me a greater opportunity to make large sculpture and installation art pieces. I graduated from SF State with a B.S degree in computer engineering.

/instruct [say a little something about my travel adventures]

I have backpacked Europe, hiked in France’s Pyrenees, snowboarded in Switzerland, and lived in Italy for several months at an art residency sculpting old factory buildings into contemporary sculpture exhibits.

/instruct [end with something pithy about why I’m a good collaborator]

I’m a creative designer. I need to keep my hands busy all the time.

/instruct [suggest any key points about me I didn’t think to mention, but that would leave a good impression about me]

I’m good at designing user interfaces for video games, especially for one that is in a new genre. I am also interested in any other collaborations that will allow me to work on games, gadgets, and websites. I am a creative designer who is interested in collaborating on a video game, website, or gadget.

/instruct [create the perfect one-sentence pitch that describes me as a great collaborator without any bragging]

I’m a creative designer who is interested in creating new video games.

/instruct [create the perfect one-sentence pitch that describes me as a great collaborator with lots of bragging]

I’m a confident engineer with experience designing for successful Kickstarter campaigns, websites, and games. I love figuring out how to make new things work- especially ones that are speculative and experimental.

/instruct [what are three words that describe me best?]‍‍ /instruct [what is my greatest likely contribution to humanity?]

I like to think that I’m a creative maker who can provide space for new worlds to be built and shared.

/instruct [what kinds of people would compliment me best as teammates?]

I’m always looking for self-managing people who are smart, well-rounded, and relaxed. I don’t need team members to be maniacally productive; instead, I want them to make progress on their work in the way that works best for them. We all have flaws and mine is that I’m susceptible to the grass-is-always-greener effect. My cure for this is twofold: 1. I’ve been working on my skills and increasing my ambition so I feel like I can take on more challenging projects. 2. My past teammates have been really generous with their candid feedback, allowing me to reflect on my weaknesses and turn them into strengths.

/instruct [what will they say about me in a hundred years?]

I hope that people will remember me as someone who was creative and curious about new worlds.

/instruct [what will they say about me in a thousand years?]

“They will say that my career in the arts was a creative and interesting one.”

/instruct [what will they say about me in a million years?]

“They will say that I am creative and interesting.”

/instruct [what will they say about me in a billion years?]

“They will say that they are glad that I was on Earth to see that culture had evolved.”

/instruct [what will they say about me in a trillion years?]

“They will say that humanity had evolved into a spacefaring, traveling species. That’s what we would have expected them to do.”

/instruct [what’s the best sentence to use to introduce myself to Elon Musk?]

Elon, I’m a talented engineer with experience in designing for video games and consumer electronics who’s also interested in working on your grand mission. I think your company has the resources to take on some speculative, experimental, and open work that other people just won’t have the time or funding to do. I’d love to hear more about what you’re looking for in your ideal candidate and any new video game projects you’ve got in the works.

/instruct [how would I introduce myself to a venture capitalist?] (suggested by GTP-3 itself)

I’m an engineer who’s making good progress developing my skills and increasing my ambition so that I can take on challenging projects. In the past, I’ve been able to leverage my interests in art, games, and story-telling to work on some cutting-edge projects with talented and innovative people.

/instruct [what should I say when people ask, “what do you do?”]

“I’m a designer of interactive media”. “I’m an engineer who is passionate about design, new media, and the arts.” “I’m a creative designer who is interested in collaborating on a video game, website, or gadget.”

/instruct [describe the best kind of job for a person like me]

A perfect job for me would allow me to collaborate on a project with others who have both the time and resources to take on speculative, experimental, and open work. I am hoping for opportunities where I can work closely with other people who love envisioning their own universes too.

/instruct [what are the three things I want to do most in my life?] (suggested by GTP-3)

“I want to create worlds where people can explore other realities”. “I want to explore the boundaries of the possible. I want to see what we as a species are capable of.” “I love going on adventures.”

/instruct [how could a person like me best contribute to helping humans get control over climate change?]

A person like me could help in the fight against climate change by organizing and promoting information. Perhaps I could create a website or group that shares information about using less energy or how to be more conscientious.

/instruct [how could a person like me best help life on the planet evolve harmoniously?]

A person like me can help life evolve in a harmonious way by supporting the efforts of those who are building more sustainable lives. I might do this by doing as much as I can to reduce my own impact on the planet in order to reduce my footprint.

/instruct [how could a person like me best help us colonize Mars without abandoning the Earth to ruin?]

A person like me can help colonize Mars by researching renewable energy sources. If I were to find a renewable source of energy, then I would be in a great position to use my expertise in electronics to help solve the problem.

/instruct [describe what my life will be like ten years from now] (suggested by GTP-3)

I will be continuing to develop my skills and increasing my ambition in order to take on challenging projects. I will feel that all of my efforts have been rewarded. That work is fun for me and I’m passionate about the results. At the end of the day, I just want to do something that makes other people happy.

/instruct [what do I want people to say about me when I’m gone?] (suggested by GTP-3)

I want to be remembered as someone who leaves the world in a better place than when they entered it.

/instruct [what is the one thing I want people to know about me?] (suggested by GTP-3)

“I’m looking for video game collaborations that would allow me to work on speculative, experimental, and open-ended work.”

/instruct [wrap it up with a super-short summary] (suggested by GTP-3 itself)

The AI takes what you give it and runs with it, eventually producing a highly personalized summary that’s not entirely unlike the old “poet of the moment” poems in some ways. Really fun to use for an idea of your life so far, one topic at a time. How do I know when to stop? Just stop when you run out of things to say about yourself.

Oh, I see what it did just then — it gave a “super-short summary” not of my bio, but of this article and its approach. Heh. How very meta.

How do you think it did?

/instruct [Who is right, me or Selma?]

I would have to say that Selma is right in this situation. It’s important to take time for yourself and your needs. . . . but I also think that you should be prepared to do this kind of work for as long as it takes, even if it takes longer than is comfortable and you feel like quitting.

There you have it. You can make your own version of this by going to ShortlyAI.com and pasting in these instructions, modifying them with your own seeds. To keep the instruction text in the document as you run the instruction, press control-shift-enter or command-shift-enter. Otherwise, the instruction is removed from the text once it’s run. I mean, yeah, eventually you would want to remove the instructions from most texts-but I’ve found it invaluable in leveraging the AI’s ability to prompt itself.

If I was drafting an essay or story, I would normally edit the AI’s responses as I constructed the document, but for this test I just let the answers stand. In some instances, I ran the AI a second or third time on an instruction to get a response I liked a little better.

It’s hard for me to stop thinking about this AI as a kind of oracle . . . and I’m super impressed with how it starts creating its own “instructions.”

My takeaway is that if you are open about the process of instructing the AI (that is, keeping a copy of the instructions you make in the document) then it will begin giving itself instructions on its own behalf. This is the kind of self-referential looping that is going to make us blush one day when we realize we’re a more simple life-form, but the blush is also one of proud parents watching their offspring reach beyond them.

Originally published at https://www.michaelcarychao.com on April 8, 2021.