My Response to TJ DeVries on AI and “Is it Still Worth it to Learn to Code”

The other day, I was watching TJ on one of his live streams. Someone in the chat asked about AI taking our jobs, and TJ responded, “I’m not convinced.” I admittedly trolled a bit and poked and prodded with more questions and statements, leading TJ to upload his take a few days later. Those of you who follow me know that I hold a degree in philosophy but worked as a software engineer at Apple and Amazon. While studying philosophy, I focused on the philosophy of mind, formal logic, and symbolic AI. I’ve been researching “AGI” for nearly a decade and even wrote my own “ChatGPT” user interface in a YouTube video with the model before what the world knows as “ChatGPT.”

Given my unique philosophy, machine learning, and engineering background, I decided to poke and prod. TJ is a talented software engineer and extremely smart in his domain. But even TJ admits in his video that he didn’t see the “ChatGPT” moment coming and didn’t think AI would be as capable as it is now. Engineers love solving problems and often must remember to question the premises of what they are trying to solve. Doing so usually leads them down rabbit holes, solving problem after problem, sometimes in the wrong direction (why companies pay for project managers). In that vein, I am not responding to TJ’s arguments or saying they are invalid, but rather that he straw-manned himself and inferred the following unsound premises.

1. AI will do “as well or better” coding tasks than junior engineers in the “next five years.” TJ makes the above claim for “argument’s sake” to find common ground. Unfortunately, he entered the wrong classroom. The notion that AI will do coding tasks as well or better than us, though possibly true, is moot. There will be no coding tasks. That is the entire point. The way we do computation is fundamentally shifting towards neural networks, symbolic AI that uses knowledge graphs, and agentic LLMs that use transformers, all of which create massive CSV files with comma-separated weight values. How the AI solved the problem in the black box doesn’t matter. The problem is solved.
2. Coding helps you learn to problem-solve and think. Though true, that isn’t a sufficient reason to put in the time and effort. Studying math, philosophy, and physics (the underlying foundation)…