My Complicated Relationship with AI

I’ve been wanting to write this post for a very long time, but haven’t been sure how to properly articulate it. But here goes…

I use AI, and I also hate AI.

I’ve been a web developer for over half my life, and in the last year or so, I’ve found myself utilizing AI tools more than I would’ve imagined. At first it was mostly as a code-completion tool. Then I got onto my first typescript project and it was a never-ending exercise in asking “what does this typescript error mean, and how do I fix it?”.

Once my editor of choice (PHPstorm) added its own AI tools, I started slowly asking it to produce code for mostly mundane and repetitive tasks. Then at some point, I was faced with having to build features for an unfamiliar platform that I didn’t really fully understand, and frankly hoped I’d never have to use again. After a lot of time spent agonizing over trying to figure it out, I decided to give AI a go at it, specifically with Claude Code.

Once I’d set up Claude to give it all the relevant context of my project, I wrote up my best understanding of what this feature needed to do and fed it to the machine. Long story short, I got at least a week’s worth of work done in about three days. And all of this working from bed with a bad back. This was the moment things started to click.

Now, did I feel bad about this? Yes, absolutely, because my brain is hard-wired for guilt and self-doubt. But I could not deny the end result. Did Claude make mistakes? Absolutely. Did I need to guide it? Sure. But what it came up with showed a deep understanding of the project that I was frankly blown away by. It’s ability to look at the existing work and intuit how things “should” be done was frankly astounding.

That was the real beginning of my journey using AI as a coding assistant. And since then, I’ve done a lot more work on honing docs and guidelines to make the process even smoother.

Sounds like I’m an AI convert, right?

Hang on.

There’s another side to AI, and that’s “generative” AI. For purposes of this article, I’m using that descriptor for anything that generates images, audio or video from prompts. This could be from Midjourney, Nano Banana or whatever else is the flavor of the week.

I’m an artist at my core. Other than a web dev, it’s the only thing I’ve ever truly wanted to be. And while I have a lot of improvement to make, I’ve worked very hard trying to get better at it.

We’re now living in a time where anyone can generate fully rendered images in seconds, by simply typing text into a box. I realize this probably feels enormously powerful and liberating for people. But I don’t think the average person using these tools understands the impact they’re having.

It’s an established fact that gen-AI is nothing but a plagiarism machine, devouring millions of pieces of real media across the internet for use in training its models. This is all without the consent of any artist. Understand that AI cannot “learn to draw”. Gen-AI uses mountains of stolen data to assemble something that has only the sheen of actual art.

You might ask, “So what, who does this really hurt?”. In the short term, maybe nobody. And your generation of a single image of a corgi riding a hippopotamus won’t exactly put an artist out of work. However, it’s the proliferation of this type of media that’s making it so huge brands like McDonalds and Coca-Cola think it’s okay to create entire commercials with AI. And in those cases, yes, positions for one or many artists will go unfilled.

I think a common refrain is “Yeah, but I can’t draw”, and that’s a pretty weak argument, in my opinion. Can’t draw? Okay, go to Youtube and find any of a thousand creator accounts that will help you learn. Practice. Have fun.

And if that’s not something you want to do, at least try to support a real artist. Even licensing a photo from Adobe or iStock would help them. Or just use a free photo from Unsplash. And there are actually ways to license the use of things like cartoons. I know not everyone can afford to commission custom artwork. And again, I’m not going to shame you for generating a couple one-off memes to share in your work Slack. But if you are using AI-generated imagery for anything resembling business purposes, I really think you’re feeding the problem, and I urge you to find ways to support real humans who have made this their life’s work.

But is all generative AI bad? Not necessarily. I can see the utility of it as a "tool" in apps like Photoshop, where it gives you the ability to stretch out backgrounds (or remove them). And I've tried a couple times to have AI generate something purely for things like perspective reference. I was looking for a specific 3/4 view of a specific type of car, and what it output was more of an amorphous, blurry blob.

Oh, the Hypocrisy!

Okay, this is the point where I acknowledge the complications in my opinions of AI. How can I laud the merits of using AI in my full-time career as a developer, yet decry it for other fields?

That’s a great question, I’d love to tell you.

In my mind, LLM’s and coding assistants are just that: assistants. They augment the skills you already have. I’m not talking about someone right out of college or high school that’s “vibe coding” sites and apps without knowing how to code it themselves. I’m speaking for experienced devs who actually do know what they’re doing, but are leveraging AI for various tasks. This could be as little as code completion, all the way to asking Claude to plan and build an entire feature.

You do still need to be a good developer. You need to know how to read the output and either finesse it by hand or ask for specific refinements. You still own the quality of what you’re shipping. A lot of times, I’ll have Claude get something started or spin up a blank component, and I’ll finish the rest. Sometimes I start from scratch and ask it to finish up a couple tedious bits.

Whereas I said earlier that AI can’t learn to draw, AI does know how to code. And even though I’m sure it benefits from the vast knowledge of years of public blog posts and Stack Overflow comments, custom apps are so unique to your project, it’s not just stealing entire codebases wholesale.

For gen-AI “art”, the model is simply saying, “Okay, they’ve asked for this collection of random things, what assets have I assimilated that I can put together to make that happen?”. The different pieces of that puzzle are individual works from artists who did not consent for them to be used that way.

I once made this argument and someone said, “But as an artist, you’ve learned to draw by referencing other art, so how is that different?”. On the surface, that seems a reasonable question. However, the difference is that I’ve spent my entire life absorbing good art from artists I love. They inspire me to explore and try to come up with my own style. Yes, I may draw a piece that’s “in the style of” Calvin & Hobbes. But I don’t claim that as my own, and I still did all the work. Someone prompting AI to do something similar and posting it as their own creation is altogether different.

It’s easy for people to counter this argument with “Well, you just can’t stand that now everyone can make art”, as if real artists are some kind of gatekeeper to an unknowable world. Everyone could always make art. Everyone. But like most things, you have to want it. You have to work for it.

So what does all this mean? Am I a hypocrite for using AI every day in my job as a developer, yet hating gen-AI with the white-hot fury of a thousand suns? Maybe. And believe me, I do struggle with this. There will always be a certain level of shame and/or guilt. But I’m confident enough in how I’m utilizing these tools to not feel like I have to hide it.

And look, I know a lot of devs don’t/won’t use AI simply because this is their “craft” and they don’t want to give that over. And believe me, I absolutely get and respect that, so I don’t think either side of that argument is wrong.

More Baggage

And I’ve not even touched on the other issues surrounding AI, such as the behemoth data centers that are using vast amounts of power and water, and having negative impacts on their communities. I do also believe that it’s costing real humans jobs, and not just artists. I really can’t imagine what it’s like as a junior developer entering this industry right now. But Claude cannot replace an entire human. It can’t interact with clients and designers to architect a solution. I’d still much rather bring on a junior dev and mentor them to master the fundamentals than have twelve LLM agents running at once.

Where does that leave me?

At this point, I’m going to continue to learn the best ways to guide and train AI development tools to be my assistants for getting things done. As far as gen-AI, I will continue to oppose and discourage its use, and call out usages where I feel it’s done a real disservice to actual artists.

I can hold two seemingly opposing beliefs in my head, while continuing to learn and evolve my position when necessary. Hopefully, I’ll append updates on this journey to this post from time to time. I'd love to hear your thoughts, whether as a developer or an artist (or both!).