Robots

As seen on my blog.

Yesterday, I read Matthew Butternick's article titled Extinction-Level Capitalism, and it has fully radicalized me. I highly recommend that you read the whole thing to have context for what I talk about in this post. In truth, I think this article forced me to look inwards and see my true beliefs. For a while now, every time I've used an LLM, I've felt a deep-seated sense of hypocrisy. I do not want to be replaced by AI, I do not want software engineers to be replaced by AI and I certainly do not want my friends to be replaced by AI. And yet, I continued to use it.

It's All a Trick (Why didn't I believe it sooner?)

Humans have made many stories about robots taking over the Earth and destroying humanity. We all know the story. What Butternick proposes is that large language models are instead a political technology. The robots are not here to kill us, physically, but instead kill us culturally. They are here to rob the working class of their right to work.

To that end, it's also a technology meant to consolidate wealth. Without any workers to pay, the the few who own and maintain the AI can slurp up more money. Of course, that is an obvious idea to basically any body with an opinion on AI, and yet we're all here living in a world with hundreds of thousands of people who use AI every day. We're all being conned into thinking that the common person can live as we were 7 years ago without LLMs dominating the topic of work.

There is always a risk with automation, a balancing act. Automation, robots, can help improve the safety and well-being of humans; however, there also still needs to be a place for humans to do the work. With AI being proposed as a tool to "increase productivity" is actually just a lie to make us rely on the technology. It's an element of the design of LLMs to fool people into thinking they are being more productive when the result of their work is simply just quantity over quality.

That's what I've realized at my job. We aren't directly told that AI is what we should be using to do our work, but there are "AI workshops" and "AI show-and-tells" where you can demonstrate how you're using AI tools to do your job. For some reason. knowing how to write good prompts is now considered a desirable skill.

A skill. That's what LLMs are being proposed as. Something you can "learn to use." That disgusts me, and I should have woken up to it earlier.

Wishy-Washy

I went into a little bit of AI-psychosis when I wrote my OpenCode post. I had never really completely vibecoded something before, that is, let an agent just write all the files for me. It bifurcated my brain, exposing two parts of me. One felt revulsion at the fact that I was doing this at all. The other felt glee at the prospect of being able to make anything instantly.

I remember expressing my disgust for AI ads, specifically Google's ads for Gemini on their phones. Then I would turn around and go to duck.ai or lumo, my perception of "the good AI." The AI that's okay to use. The ones that are okay to replace your thinking with.

I've used AI to do school work in college. I've used it to write parts of my own coding projects, though I tried to manually type out as much of it as I could. My knowledge of TypeScript and JavaScript are filled with so many holes because I learned both languages around the time when Copilot was added to Visual Studio Code.

I could feel, at the time, that AI was an easy-button, that I shouldn't rely on it too much. I had read Dune well before all of this started happening. I was revolted by the blatant greed of the companies making LLMs and this new term called vibecoding that had been floating around. But I used Copilot anyways.

Brain Surgery

We all make mistakes. That's what humans do. My mistake was several years long, and that was trusting LLMs with my brain. Now, I'm taking it back. I'm never going to put an error into an AI again. I'm never going to prompt an AI ever again. I will never watch as the little spinner spins away next to pixels on the screen that spell "Thinking..." LLMs don't think, humans do. I am human.

My company might try to make me use AI, but I don't want to. My coworkers might find me strange for quickly opposing any use of AI, especially since I've used it in the past, even participated in one of those AI show-and-tells, but my mind is no longer bifurcated. I'm going to break this habit of using LLMs to do my work.

Unfortunately, there's no way you can hold me accountable. My mind could still be bifurcated and you would never know. But I think it's going to be easier to do this than I think. I'm already enjoying to process of untangling my vibecoded project at work. My mind is tantalized at the thought of pushing code to a Codeberg repository that was truly written by me.

Revolt

I am against Flock, data centers and LLMs. Our property and our minds should not be trampled by the schemes of the wealthy. We do not need to ask their machine gods for help because we have all the help that we need, and it's in our minds and in each other. Technology is meant to help humans, not replace them. Robots are meant to save us time, not work.

Work means the drawings, documentation, code, creative writing, whatever is made by the human mind. Sometimes making that stuff is fun, sometimes it's excruciating, but it's all uniquely human. The only reason LLMs are being pushed on the worker is so that work can be replaced, so that the system we have all grown to rely on can be uprooted. Humans must continue to work, to produce the mistakes and flaws and the beautiful work that permeates the entire story of our civilization.

Our world is the sum of the work our ancestors did and all of the work we do together is why we still have a society. That is the work the AI companies want to replace. To offload work to LLMs is to effectively deny the usefulness of anything a human could produce.

Be Butlerian

In the universe of Dune, there was an event called the Butlerian Jihad. People across the galaxy revolted and destroyed all the "thinking machines" they could find. It was a story device that Frank Herbert, the author of Dune, used to make the book about people and politics rather than technology, something unexpected of sci-fi at the time. People still use technology in that universe, it's just that none of it "thinks." The human mind is the only matter in that universe that thinks.

We lived in that world for a while. We had machine learning, algorithms capable of automatically detecting patterns in text or images and associating those patterns with a single word, but everything was written by a human. There was no "slop" flooding the internet. But now we have thinking machines. The massive data centers doing the thinking for us, polluting our physical and digital spaces at the same time.

We should continue to use machines. They are helpful; my thoughts would not be able to reach you as they are now without machines. I like my computer, a machine. In fact, I recently named it basalt. But it does not think. I think of ideas and express them to it. We work in tandem. Do not let the popular promise of artificial intelligence take root in your mind; it's a lie to make you complacent. That, at least, is what I wish I could say to my past self.

Good-bye AI

Hopefully this is the last article I write about AI. I'm sure you're sick of hearing about it, and I'm sick of hearing about vibecoding. In the future, you can expect to see 100% human-written posts on this website about what was on my mind during the day.

Maybe if I stop reading all of the posts tagged vibecoding on lobsters I won't think about AI so much...