AI

I have very mixed feelings about “AI”. Or, more accurately, I have very mixed feelings about what people tend to mean when they refer to AI.

The field of AI is vast and has been around since the 1950s. These days, when someone says AI they’re most likely referring to LLMs. For the remainder of this page, I will largely be referring to LLMs when I say “AI”.

Problems with AI

I want to start this discussion by acknowledging that there are many problems with AI.

  • Modern LLMs were built by unethically stealing the work of many millions of people without gaining any sort of consent or providing any kind of compensation.
  • This theft of data continues with complete disregard to the privacy laws that are supposed to protect use from ths sort of thing.
  • The technology is being used to directly harm people. Everything from encouraging harmful behavior to minors, rapidly accelerating disinformation, and even producing deep-fake videos of non-consenting individuals.
  • AI makes all writing that utilizes it feel flat, boring, and completely devoid of any human characteristics.
  • It completely devalues labor in a large number of categories to enrich the ultra wealthy even more than they already are.
  • LLMs use an unimaginable amount of power and resources. The amount of harm being done by this point alone is staggering.
  • This technology can actively atrophy our ability to be creative and think critically.
  • This technology can harm our social relationships.

Overall, I have a great deal of concern over where we’ll end up with all of these problems. It’s my hope that the current hype-bubble will eventually pop. When it does, I truly hope that we end up in a world where AI can be responsibly be used in the areas that it’s actually useful and helpful without it being used to continue harming us.

Unfortunately, I tend to lean towards the cynical side, which means that I don’t think this is where we’ll actually end up.

When I absolutely do not use AI

I want ot start by mentioning the times that I will absolutely never use AI:

In summary, I absolutely refuse to use AI for any of my creative endeavors. It is my very strong opinion that AI has absolutely no place in anything classified as art. AI “art” is not art.

When I might use AI

Despite being largely opposed to AI in most areas of my life, there are instances where I might use AI. Well, in reality, there are exactly two instances where I might use AI:

  1. Helping me search for information. One thing LLMs are great at is helping you surface information and finding related sources for that information. With the continued enshitification of services like Google, this is a much preferred way to find information as compared to using a search engine.
  2. Software development. This one is a slippery slope, but there’s a huge difference between using LLMs responsibly for software development and producing AI slop.

AI use with software development

As I mentioned, there’s a huge difference between “vibe coding” and doing actual software development.

I’m a software engineer by profession and I have a degree in computer science. I have spent many years building up my knowledge of software development and how computers work. As useful as LLMs can be for software development, they will never fully replace a knowledgeable human. “Vibe coding” without understanding what you’re doing and how things work will lead to a massive pile of AI slop. I don’t want to hear anyone’s arguments about how this isn’t true and how AI can “one shot” whatever you throw at it. It’s simply bullshit.

So, where do I stand with AI-assisted software development? Well, I’m actually a fan of it. This is one of the few areas where I think LLMs are truly useful. It has allowed me to greatly accelerate how fast I can move and has allowed me to start exploring side projects that I would’ve otherwise never had the time to explore.

With all of this said, I have a series of principles I follow when utilizing AI for software development:

  • I never outsource all of my thinking to AI. I will use it to help me implement an idea, but the AI does not do the thinking for me.
  • I never let the AI make the architectural decisions.
  • I always read the code and make sure I understand it. If I don’t understand the code the AI produces, then the code is not good enough to commit to the project. Either the code is bad or my understanding of the problem set is too limited. In either case, it would be irresponsible to commit that code.
  • I never “vibe code” and don’t produce slop. I am a professional and I take a great deal of pride in my craft.

I also don’t always utilize LLMs for software development. I truly enjoy the art of programming. As such, there’s plenty of times that I simply don’t bother using an LLM to assist me and just produce all of the code myself. Not only is this enjoyable, it prevents my skills from atrophying.

This site is completely AI free

I do not use AI to generate the content on this site. I don’t use AI to write my articles. I don’t use AI to generate any of my artwork. While I do use AI to assist with some research and some development, I do not rely on it and always verify everything I use it for.