I am not an artist. At least, not in the traditional sense. I have some background in media production, and I’ve spent enough time working with graphics, video, and websites to know what I like and recognize when something isn’t working. But if you handed me a blank canvas and asked me to illustrate a duck holding a T-shirt, we would both be disappointed with the results.
I’m also not a programmer. I know enough HTML to modify a webpage and enough CSS to occasionally make things worse. I can follow code, recognize certain patterns, and usually figure out what I’m looking at, but nobody should be hiring me to build their next application.
And yet, recently, I worked with ChatGPT to produce an accessible Brightspace announcement, develop some custom HTML and CSS, and create a banner illustration for the article I wrote about the experience.
I made those things, but I didn’t make them alone. And that distinction has been rattling around in my head ever since.
What does it mean to make something?
In my previous article, I described how a seemingly simple registration announcement turned into a much longer conversation about accessibility, screen readers, HTML, CSS, and the process of learning through AI-assisted problem-solving. The original task was to post a flyer announcing Spring 2027 registration at College of the Mainland. By the end, I had learned about visually hidden text, questioned several recommendations from ChatGPT, and used Chrome’s developer tools to investigate whether our solution was working as intended.
Then I needed a banner image for the article.
I could have grabbed a stock photo of somebody typing on a laptop, preferably with those familiar glowing blue technology symbols floating around their head. We apparently can’t discuss technology without somebody staring thoughtfully at a hologram.
Instead, I worked with AI to create an illustration that connected the story’s elements: a duck, a computer, accessibility, and the free T-shirt that started the whole thing.
The first version was too busy. It had more visual information than I wanted, and I knew SharePoint would crop it unpredictably across different screen sizes. So I explained the problem and asked for something simpler. We eventually arrived at a cleaner illustration that worked better for the space.
I made decisions about the composition, subject, purpose, and presentation, but I didn’t draw the illustration. The AI generated the pixels. Both statements are true.
Would I call myself the artist? Probably not. Would I say I had nothing to do with creating it? That doesn’t seem accurate either.
I’ve been thinking about where we draw that line, particularly when the discussion turns to AI-generated images, writing, music, and software.
We already accept many forms of assisted creation without questioning whether the person using the tool contributed anything meaningful. A photographer doesn’t have to manufacture a camera. A filmmaker relies on cameras, editing software, sound equipment, and the contributions of numerous other people. Digital artists use tools that automate processes that once required considerable manual effort.
Those comparisons aren’t perfect. Generative AI can contribute expressive decisions in ways that a camera or a paintbrush generally doesn’t, and that difference deserves serious consideration. But it also seems unreasonable to pretend that every use of these systems involves the same degree of human involvement.
There’s a meaningful difference between typing “make me something cool,” accepting whatever appears, and spending time developing, evaluating, revising, and integrating an output into a larger work.
That doesn’t automatically make the second person an artist. It does suggest that the human contribution is worth examining.
Lowering the barrier doesn’t eliminate the responsibility
One of the more interesting developments around large language models is how they’re changing who can attempt certain technical tasks.
We’ve seen the rise of vibe coding, where people describe what they want software to do in ordinary language and use AI tools to generate much of the underlying code. Someone who has never studied programming can now experiment with building a simple website, automating a repetitive process, or creating a working prototype.
That’s exciting, particularly in education. There are faculty members with wonderful instructional ideas who may lack the technical skills or resources to build the tools they imagine. Giving those people a way to experiment could open possibilities that previously remained stuck in the idea stage.
It also creates problems.
A working application isn’t necessarily a secure application, and a person who doesn’t understand the underlying code may struggle to recognize vulnerabilities or know when something is unsafe to deploy. Researchers studying AI-assisted software development have already identified serious security concerns in applications created with these approaches.
The same capability that helps someone create a useful tool can also help someone create something harmful, whether intentionally or through ignorance.
We’re seeing similar tensions in AI-assisted reverse engineering. Tools are becoming more capable of helping people examine software, understand how systems function, and identify vulnerabilities. For security researchers, accessibility developers, and people maintaining older systems, those capabilities can be extremely useful. They can also be exploited by people looking for weaknesses they have no intention of fixing.
The technology doesn’t make those intentions equivalent, and we shouldn’t pretend that beneficial uses excuse malicious ones. There are legitimate concerns about security, ownership, privacy, and the ability to scale harmful activity.
But I also don’t think the existence of bad actors means every person using these tools should be treated as one.
A faculty member using an LLM to understand unfamiliar HTML is doing something fundamentally different from someone using the same technology to compromise a system. A person developing an illustration through an extended creative process is not necessarily doing the same thing as someone flooding social media with thousands of misleading images.
We need to be capable of recognizing those differences without dismissing the risks that come with the technology.
The problem with calling everything slop
I’m not particularly interested in defending AI slop.
We’ve all encountered it. Generic articles that say very little in far too many words. Images with inconsistent details. Videos designed to capture attention without offering much substance. Content produced in enormous quantities because generating it is easier than thinking about whether anyone needs it.
The criticism is deserved, and in many cases, the tools have made an existing problem considerably worse. The internet wasn’t exactly suffering from a shortage of low-quality content before generative AI arrived.
What bothers me is how easily the criticism can shift from evaluating the work to dismissing the process.
If someone says an illustration is poorly composed, derivative, misleading, or inappropriate for its purpose, that’s a criticism we can discuss. If an article contains fabricated information, weak reasoning, or no meaningful human contribution, those are problems worth identifying.
But simply discovering that an LLM was involved doesn’t tell us everything we need to know.
Consider the Brightspace announcement I worked on. An LLM helped me understand accessibility techniques, suggested HTML, and explained CSS properties I didn’t fully understand. I made the decisions about what students needed, rejected unnecessary content, modified the implementation, and investigated the result.
Would the final announcement somehow be more valuable if I’d spent several hours searching through forums and documentation instead of having a conversation with ChatGPT?
I would still need to verify the information. I would still need to understand the code well enough to use it responsibly. And I would still be accountable for the experience students had when they encountered the announcement.
The important questions seem to have more to do with those decisions than with whether an LLM participated.
The illustration raises a more complicated question because the system contributed much more of the finished visual expression. That’s a legitimate distinction, particularly when discussing artistic skill, intellectual property, and creative labor. The U.S. Copyright Office has recognized that human creative contributions can be protected in AI-assisted works, while simply providing prompts does not automatically establish copyright over the generated material.
That isn’t a universal answer to the question of authorship, but it reflects something worth considering: human involvement exists on a spectrum, and the nature of that involvement matters.
I’m not asking anyone to admire a generated image simply because I spent time refining it. I’m asking whether we can evaluate the image, understand the process, and acknowledge the human contribution without pretending the software did nothing or the human did everything.
Can we stop calling it artificial intelligence yet?
I’ve complained about this before, and I suspect I’ll continue complaining long after everyone has stopped listening.
I don’t particularly like the term artificial intelligence, especially when discussing large language models. It encourages people to imagine a thinking entity behind the interface, which can lead to some fairly questionable assumptions about what these systems understand and how they operate.
I’d prefer that we use more precise language when discussing what the technology actually does.
Unfortunately, I think I’ve lost that battle.
We’re probably at the point where AI has become one of those words that escaped its original meaning and settled into everyday language. People ask for a Coke when they mean a soft drink, even though Coca-Cola is a particular brand. Not everyone does that, and I imagine a few beverage companies would appreciate the distinction, but language has a habit of doing whatever it wants.
So here I am, once again trying to take back a term that has already moved into the neighborhood, bought a house, and started receiving mail.
Still, I think the distinction matters when we talk about creation.
ChatGPT isn’t human. It doesn’t have my experiences, my responsibilities, or my reasons for wanting to make an announcement accessible. It can generate a response, explain an approach, or help produce an image, but it doesn’t share the human experience behind why I wanted those things in the first place.
When I describe it as an assistant, I’m using a practical analogy for the role it plays in my work, not suggesting that there’s another person sitting at the keyboard.
I bring the context, goals, preferences, and responsibility. The system provides capabilities I can use, question, and direct. Sometimes those capabilities are impressive. Sometimes they produce complete nonsense. Occasionally they manage both in the same conversation.
The relationship is useful precisely because I don’t have to pretend the tool is human to benefit from interacting with it.
What are we actually teaching students to value?
For educators, I think this conversation eventually comes back to what we’re asking students to demonstrate.
If an assignment is intended to assess a student’s ability to write code independently, submitting something generated by an LLM without permission or understanding is a problem. If we’re assessing illustration techniques, having a model produce the artwork doesn’t demonstrate those techniques.
But not every assignment is designed to evaluate the same skills.
Sometimes we’re assessing whether a student can solve a problem, communicate effectively, evaluate information, develop an instructional resource, or make informed decisions about technology. In those situations, the use of assistance may be relevant without automatically invalidating the work.
The challenge is defining what the student is responsible for and how they demonstrate their contribution.
We can ask students to explain their decisions, document their process, identify errors, justify revisions, and demonstrate understanding beyond the generated output. We can also establish circumstances where AI assistance isn’t appropriate because the skill being assessed requires independent performance.
That seems more productive than treating every AI-assisted submission as evidence of laziness or treating every generated product as an achievement.
There are also questions of equity. Not everyone arrives with the same technical skills, access to equipment, or opportunities to learn creative tools. An LLM may help someone explore an idea that would otherwise be inaccessible to them. That possibility deserves consideration alongside concerns about dependence, fairness, and the potential erosion of foundational skills.
I don’t want students to stop learning to write, code, draw, or think because a machine can produce something that resembles the finished product. I also don’t want to assume that the only meaningful learning happens when someone completes every part of a task without assistance.
The distinction depends on what we’re trying to teach.
I still made something
The registration announcement is a fairly ordinary piece of work. Students will see it, hopefully register for classes, and perhaps pick up a free T-shirt. The banner illustration is a small graphic attached to a blog article. Neither is going to change the world.
But creating them gave me an opportunity to learn, experiment, and produce something I couldn’t have made in quite the same way on my own.
I didn’t become a programmer because I used AI-generated CSS. I didn’t become an illustrator because I directed the creation of a graphic. And the LLM didn’t become a person because we worked through a problem conversationally.
What happened was considerably less dramatic. I had an idea, encountered limitations in my own skills, used a tool to extend what I could do, and made decisions about the result.
Some of those decisions were good. Others needed revision. I had to question the tool, evaluate its suggestions, and remain responsible for what I published.
That’s a process I recognize from plenty of creative and technical work that has nothing to do with AI.
We should absolutely criticize low-effort content, challenge unethical practices, protect creative work, and hold people accountable for the things they publish. The technology deserves scrutiny, particularly as it becomes more capable and more widely available.
I just hope we can leave enough room in that conversation to recognize when someone has used these tools thoughtfully to make something useful, interesting, or meaningful.
I am not an artist, and I made that. With help, obviously.
