About LLMs, Star Wars and Sovereignty

It looks like my profession, Software development, will be among the industries which are most affected by the rise of llms. The pace of progress has been kind of mindboggling. From using a chat window to ask ChatGPT stuff, via using a VSCode Github Copilot plugin and finally reaching the Claude CLI or Gemini CLI or Codex endgame.

It's staggering, the amount of changes to workflows I have seen recently.

The 'Star Wars' future

In recent months I am often reminded of a certain paradox I (and probably others) noted in Star Wars: "Why do the user interfaces of their ships and computers look like they are from the 70s or 80s?". The realistic answer is obvious, because the films were made back then! But what if the in-universe answer is that machines like LLMs started to exist back when that look was cutting edge. What if they were good enough to produce any kind of software the galaxy needed. There would be no need to invent something further for so long and, by the time that it MIGHT be needed, no one knew how to change it anymore, since the LLMs did all the programming for generations now. It's basically lost knowledge. In our world, that clunky 70s UX design, might be React. React forever, all the way down!

The Issue with Sovereignty

I'm from the EU. I would call myself a somewhat proud European. The EU, in some way, might be the last best hope for humanity. Up until now, we tried not to participate too much in the race to the bottom of ever longer working hours. However, for a couple of years and decades now, this hope is being dashed. From higher up the technological ladder by the Americans, from further down the industrial ladder (at least, in parts) by China. And, most recently, militarily by Russia.

If America wanted to, they could probably shut down a very high double digit percentage of our economy with a couple of pen strokes. AWS, bricking Windows updates, Google Cloud, Cloudflare. I could go on. And now, to that ever growing lists of dependencies, the LLMs have been added.

My guess would be that a not insignificant part of software development in the EU would grind to a halt without access to Anthropic, OpenAI or Google. European offerings, especially in regards to SOTA coding, are limited to none existent. If we take the silicon itself into the mix, then everything would stop. There are only fledging EU-native SoC designs and no SOTA fab as far as I know. (And yes, I know about that one firm)

Personal Sovereignty

I, for one, will try to keep my none LLM coding skills at least somewhat sharp. In certain environments especially, LLMs are nothing more than a security vulnerability. Not only since they might produce cleary bad, or worse, subtly bad code. But also to be able to function during times when they are not available. Either from a simple outage, or a large scale economic or kinetic war.