When is Using Generative AI "Icky"?

Note: the contents of this blog post have not been written, edited, or proofread by AI. I also did not use AI to research anything in writing this post, and the opinions are all mine. Any reference to "AI" in this post can be interpreted as generative AI specifically.

Among my peers, generative AI is a pretty controversial thing. I have classmates, professors, and colleagues who have sworn not to use it at all, but also some that are vibe coding every day without a second thought.

I use Gen AI, and I use it pretty often. Sometimes I probably use it more than I should. But I also have strong opinions on when it is and isn't okay to use it, and you should too, even if they're a little different from mine.

"Icky" Explained

Writing this blog post using AI would be icky. Generating media assets such as images, video, music, illustrations, logos, etc.? All of that is icky. Educational materials created using AI? Quite icky to me.

On the other hand: using AI to help me organize my job search? Not icky to me. Asking Claude to debug some code that I wrote myself? Usually not icky.

Just to be clear, the "icky" and "not icky" distinction has nothing to do with the serious social and economic issues created by generative AI nor the unethical things being done by AI companies. It is very valid to decide that you do not want to use AI at all because of those things.

Instead, this idea of "icky" is about having discipline if/when you do want to use AI. If you already use AI frequently, it's a way of analyzing whether you're using it in a way that is benefitting you instead of hurting you.

My Icky-ness Table

Every person has different opinions on what is icky, but here's a table describing some of my general thoughts:

In my opinion, any user of generative AI should have an at least semi-concrete definition of what is "icky" to them. It doesn't have to be set in stone or complete, but it should be something we are always thinking about.

So...how do I actually decide what is icky?

Icky-ness will always depend on the context and your own opinions, but here are a few questions you can ask yourself to determine if using AI feels icky to you:

Does it undermine my learning?

If you would learn a significant amount of valuable information by completing a task yourself, you should probably do it yourself. Of course, "valuable" is a subjective word, and it's up to you to decide what that means, but if your goal is to learn, then you shouldn't be automating the thing you want to learn.

If anything, your uses of AI should aim to aid learning. I'm not against using it to briefly summarize a new topic in a digestible way, or to ask clarifying questions, I think it's great at that. You can use it to get you unstuck when you're confused about a new topic, but keep in mind that struggle is often a better teacher than success. AI can give you the information, but it can't synthesize it into your brain for you. (yet?)

Should it be personal?

A good rule of thumb is that if an end result should reflect the personality of its creator, or represent the personality of a company or brand, then it should probably be created by a human. In any discipline of art, the artist's subtle decisions build up over time and lead to a very personal and unique piece. This is why, even though LLMs are literally designed to read and write text, I actually find writing text with AI to be very icky. In real writing, every single word and syntax choice is a subconscious reflection of the author, so to write with AI feels deeply impersonal to me.

Who is the Audience?

Apologies to every tech startup right now: selling an entirely vibe coded product for money is icky. Generating class materials that are given out to students is icky. As a consumer or student, I would feel like I've been cheated (in reality, this has happened to me multiple times, and I do feel cheated). On the other hand, vibe coding a project that is useful or fun that you will never release feels fine to me. It also feels fine to vibe code software and then open-source it publicly. If you choose to entirely vibe code something, it's very icky to then put that on a portfolio or resume for a recruiter to perceive as entirely your work.

How much of the final result will be mine?

Are you vibe coding entirely from scratch, or just asking AI to debug your existing code? Generally, the less you use AI, the less icky it can realistically be. The more you use AI, the more you should be retroactively examining what it does. At the very least, that lets you learn from what it did. It's easy to go down a rabbit hole of using AI for one thing, then realizing it could do this other thing also, and surely it could add this new feature pretty easily. The more you do this, the less you learn and the less personal your final result will be.

Closing Thoughts

For me, "icky" is just an abstract metric of how okay or not-okay it is, from a personal perspective, to use generative AI for a task. The goal is to decide if using AI in some way will actually help you in the long run. In a world where it kind of feels like I need to be using AI to stay ahead, and the practicality of it makes me want to, it's a way of deciding exactly when I actually should.

TL;DR: Before you use generative AI, simply ask yourself if it feels icky.