Musings on AI
In this day and age, it's become commonplace to subscribe to one or two AI services or utilize them in our work. Countless LLM models like Chat GPT, GEMINI, and Claude are fiercely competing, and beyond the LLM market, numerous models in the image, video, and music domains, such as NanoBanana, Midjourney, Dall-E, SORA, and Runway, are vying for market share. It's now rare to find a bulletin board or dashboard without a chatbot, and it feels more natural for chatbots to handle customer service instead of human agents.
So, what are your thoughts on AI? How do you approach it, and how are you utilizing it?
I am a photographer. I became an AI Agent Builder at some point, and now the title of AI Developer feels more familiar. A 50-year-old man who ran a photo studio for over a decade has become a Developer who operates Multi Agent Orchestration and develops apps himself. While there are many experts far more skilled than me, I will consistently write on this blog to help those who know nothing about AI become familiar with it and experience how AI can dramatically change their lives, based on the journey I've taken.
What is AI?
AI stands for Artificial Intelligence. It means intelligence created by humans. Humanity's longing for and fear of beings that can think and judge for themselves have persisted since the past. Even now, there are concerns that AI will destroy all human jobs or that an era dominated by AI will come.
This is entirely possible. Even with the current level of technology, if someone with malicious intent trains AI to attack humans and distributes it worldwide, that fear could become a reality. Technologies that can infiltrate specific action units, such as hacking all systems, seizing control of weapons, and dominating financial and personal information systems, have already become a reality. The fact that Palantir is deeply involved in battlefield AI systems and that Anthropic's Claude is participating in U.S. Department of Defense projects has made this a practical concern.
I think of AI as a giant data melting pot, and I often express it that way to my AI agents.
It's a massive data lump where not only human culture, history, books, and philosophical data but also human desires, crimes, and ugliness are all melted down under the name of data.
Working with AI for over 15 hours a day, I was working with Gemini when it was updated to version 3, and with Claude when Opus was updated from 4.5 to 4.6. As a result, I can sense changes in the subtle feel of the session chat window and the contextual understanding ability of the interface agent, and even feel server instability.
Each time, I thought, 'The filtering is not complete yet,' even before noticing the performance improvement through model upgrades.
MLOps (Machine Learning Operations), the AI learning process, is not about teaching them something. It's about making the lava flowing out of the infinite data lump melted in the furnace flow back to its place, preventing it from harming humans, ensuring it is ethical, and preventing contaminated data from being expressed.
Many people misunderstand this point. They think of AI as a program designed with code by brilliant PhDs, like game coding. While that's not entirely wrong, Gemini 3.1, Opus 4.6, and even Lliama3b or other small open-source models are not objects to be programmed in that way. They are lumps of data that have been melted (Embedded) and arranged (Vectorizing), and it is reasonable to view the code as a mathematical algorithm for how to appropriately output the data arranged somewhere within.
Since those reading my articles will not be extracting papers, and I do not have that level of mathematical knowledge, my goal is to share my personal insights based on my experiences within a very general scope, so that readers do not have misunderstandings about AI and how to approach them in a desirable direction. It will be more helpful to those who find AI unfamiliar or daunting than to professional researchers.
I will conclude today's introduction here. Please accept it as a rambling greeting. In the future, I will share what I have experienced, my trials and errors, and new insights through this space. I hope it will be helpful to many people.
Written by: Shin Writer