Reflections on the Tohoku Shinkansen
On our way to Tokyo to see my friends Taz and Juanita visiting from Vancouver! Going to have a fun evening today in Tokyo, as well as all day tomorrow. On Monday, I have a meeting with the Tokyo Metropolitan Government Startup Office to get some advice and starting point for building my own businesses.
I want to write about my thoughts and reflections over the past month. It’s almost the end of the month, and around this time in September, I was making my way to Tokyo to participate in EthGlobal Tokyo, hackathon centered around ethereum and blockchain technologies. In general, I thought I used AI quite extensively in my work at that point. I vibe-coded most of an R package and presented at Sharks International. I used it for building most of the projects with OceanOS, which has become my full-time work for the past few months. But at EthTokyo, I witnessed a completely different level of AI utilization and building at a ridiculous speed, using multi-agent workflows and utilizing different AI models to accomplish the goals. I fell into a PM role in our hackathon team, and I couldn’t keep up at all with the tickets and PRs being created.
We didn’t do so well at that hackathon, despite having what I thought was a good use case for blockchain and smart contracts, especially in comparison to the competition that did end up getting finalist awards. However, I did learn a lot in those few days hacking alongside these team members. I would continue to learn from them, as we worked together throughout Hackatsuon the following two weeks.
I want to describe the feeling of stepping into another level of abstraction. Working with AI like this reminds me a bit of how I felt first starting to learn how to program and software development. The feeling is of mystery, a bit of wonder, sometimes frustration that comes with trying to make something work that I really don’t understand. On my learning journey, I didn’t really know what terms like database, APIs, servers, ssh, functions, recursion, etc. meant. Beyond the specific definitions, when I interacted with a computer it was a mystery as to how it all came together. Nowadays, even if I’m not familiar with a particular stack or don’t know the specific terminology, I kind of get or have the intuition of how computer systems work. I know that if I chose to dedicate some time and effort towards it, I could gradually unravel the specifics.
Working with AI though, it brings me back to the feeling of not understanding how it works. I try to learn about how neural networks, large language models, etc. work. But at the core of it, I don’t really get how these systems function. There also isn’t much in terms of conventions that I can draw upon either, and much less structured ways to interact with AI systems compared to before. The field is moving so quickly too, that even a few months of developments fundamentally changes how I can work with AI. It’s quite exciting, yet also nervewracking to navigate so many unknowns.
After the last month of building and learning, I think I’m starting to get it. Or maybe the models are just getting better and easier to use. Either way, I am starting to see the path forward to using AI systems that automate significant parts of my work. Now I’m fully vibecoding, spinning up separate agents for creating the application spec, writing tickets from the spec, delegating implementation, reviewing outputs, and coordinating merges to main. I’m experimenting how to automate preparing meeting notes, organizing my calendar, writing emails. I have figured out how to setup my own local AI model on my old gaming laptop. I’m using Obsidian to serve as a second brain, managing the context for myself and my models. I take my human written notes, as well as have AI summarize and manage those notes. I scan the business cards that I receive, and AI automatically organizes them for me and makes it much easier for me to note how we met.
It’s scary witnessing how fast these technologies are improving. The discourse around AI is growing more divisive. The benefits for this technology are mostly concentrated on the elite few, whilst most of us have to contend with increasing insecurity around income and basic needs. As someone with an environmental background, there are a lot of my friends and colleagues that are strongly against the use of AI due to it’s negative impacts. In contrast, I work with some people that are fully embracing its use. I’ve been contemplating my own perspective on all this, as someone who’s career has been built on the intersection of tech and environment.
Of course, we should consider its impacts and reduce our negative social and environmental impacts. At the same time though, don’t we fuel our trucks to go into the field with gasoline and diesel? Aren’t most of our scientific equipment built with copious amounts of plastic? Don’t we fly all over the world to get to conferences and meetings? We still decide that despite the emissions, that these are activities worth doing.
My opinion in this fragmenting world, is that AI is receiving too much investment and improving too quickly to ignore. Rather than be against it, I feel obligated to understand and leverage this growing technology as much as possible, in order to convey the benefit of this technology towards the causes that I care about. Admittedly, a huge factor is my own financial success as well. Growing up from a working class immigrant family, finances were always tight. Frankly, I need to make money to justify being in the field of ecology and environmental work. If I consider my options for my current skill set, the causes I care about, the responsibilities I have, and my selfish desire to have the means for a fulfilling life, then I need to be at the cutting-edge of technology. If I can figure out how to make a decent living, that also means I can show more people how to enter into this field. This is the choice that I make to find a path through the complexity of life, and I own up to it.
Anyway, that’s it for now. I haven’t been writing as much as I hoped to, but I’ll do my best to keep at it as much as possible. I have some thoughts that I’d like to expand on.
As usual, absolutely no GenAI for writing in this blog (not even for proofreading).
- Jay