It’s useful to craft a personal narrative. At least professionally. For oneself and for others. It might not be as clear as it appears, but it is still helpful. This is mine. ## Early days I grew up in Mexico City. I love it there. I was never much of a traditional geek: most of my childhood was spent playing outside. I learned many instruments, played in questionably sounding bands and dabbled into almost any sport imaginable. I also loved computers. I remember the first time I used Google. I was blown away. I couldn’t believe I could get infinite images of Spider Man just by typing in a text box. Yes, that was the first thing I Googled. I remember it vividly. Sometime in my teenage years I became obsessed with "making a change". Mostly I wanted to improve my native Mexico. So I decided to study economics at university. I thought economic structures were the most important thing shaping society. I found the maths of human behaviour fascinating. But then I became more and more interested in AI and technology. First, because it felt like an endless playground for my creativity. I could create impactful things by writing code. Second, because I read about AI. I realised this would be the most transformational thing in my lifetime. Yes, economic structures shape society. But AI would shape economic structures like nothing before. Perhaps most importantly, I was just fascinated by it. I decided to switch majors to Computer Engineering. ## University years During my degree, I wanted to learn more about data science and machine learning than I was learning in class, and figured the best way was to work on real-world projects. So I founded a student-led organisation that worked on data science problems for social good. The idea was simple: multidisciplinary students learn data skills together. We then help NGO's with data problems in the real world. We learn more because of it. A win-win. Along with my co-founder and best friend, we initially got 20 students to participate. We started teaching ourselves collectively about data science and machine learning on Saturdays. We started approaching NGO's and then they started approaching us. We worked on some cool stuff, including a nationally touring art installation for a missing persons organisation. This is when I first became interested in the creative potential of AI. ## After uni, Google and Australia During my last year I joined Google as an intern in the Mexico City office. Then I joined full-time while still finishing my degree. I worked mostly on data and ads stuff. The company was amazing, people were incredible, but I didn’t find the job very interesting. I gradually became more interested in two things: the social implications of AI and using AI as a creative tool in weird ways: to make music, images and write things. One day, scrolling Twitter, I saw a call from one of my favourite follows: Genevieve Bell. She’s an anthropologist who somehow became a technologist and VP at Intel in Silicon Valley, exploring the anthropology of technology. She was coming back to Australia, where she is originally from, to lead a new institute exploring AI and technology from an interdisciplinary lens and was calling people from all over the world to join, offering a full scholarship. I didn’t think twice. ## Writing a master's application with AI? Bear with me. As part of the application I had to write an essay about why I wanted to join. I wrote and rewrote drafts for many months, never satisfied. One day, as I was at an AI conference, I had an idea. I had just heard about GPT2. A language model by OpenAI. This was 2019. It was open source. You could download it and run it in your computer to generate text! This was ground breaking stuff then. So I left the conference and went home. My idea was to co-write the essay with this AI model: I would write one paragraph, then generate the next one with the model and so on, iteratively. The essay was about the implications of human-AI collaboration. This was a great idea. But there was one day left to submit the essay. I battled all night to install GPT2 locally and wrangle some code to get it to spit out text. Kids these days don’t know how good they have it with ChatGPT. I literally didn’t sleep that night. I eventually got a Jupyter notebook working that somehow seemlessly let me co-write with this AI partner. I made the essay and submitted it. I got accepted. Packed my bags (and by bike), and came to Australia. ## Australia I arrived in 2020. Two weeks before the pandemic. Everything closed. But I lived in a little studio by a gorgeous lake and I had my bike. I didn’t need much else. The masters was an unbelievable experience. I became fascinated by complex systems, cybernetics, evolutionary systems and distributed intelligence. These expanded my thinking around intelligent systems in a way that I would have never anticipated. I became enamoured by Australian nature, and I spent as much time as a could outside. This is when I became interested in the intersection between technology and nature, and how AI could be applied to care for it. Perhaps my greatest achievement during the masters was creating a smart bird house that identified, monitored and fed bird species. The VC of the university and Nobel Prize winner Brian Schmidt saw it at the masters demo day and told me he wanted one for his house. ## Stay in Australia or go back? Around this time, by the end of my PhD, I started working with the CSIRO (the Australian Science Agency) for my final project. With them I developed a new algorithm for invasive species management, one of Australia’s biggest ecological problems. I was then hired full-time. I originally came for only a year and a half to Australia. But it started to seem like I would stay longer. Someone from my masters sent me yet another tweet, this time from Oliver Bown, a professor looking for PhD students wanting to research human-AI creative collaboration. So I moved to Sydney. This was 2021. Right before AI really took off. ## Narrative Device I’ve been fascinated by language models since I first got to play with GPT2. A rogue, uncensored and wild version of what we have today. It was cool, but hard to make useful things with it. But 2022, InstructGPT was launched. It was much better at following instructions and generally at writing creative and cohesive text. I had an idea: can it write coherent things about wildly different things? I wrote a little app called Narrative Device which took two words, any words, and spit out a short story combining both concepts. This thing was a hit, literally overnight. I can say that for thousands of people, this was the first time they saw what a language model was capable of. More than 2,000,000 stories were generated. I exhausted my OpenAI spend limit, many times. People donated, they kept it running. Eventually, the hype cycle died off, but I learned many things. Which I will write about shortly. ## My PhD in Sydney (in progress...)