There are different levels to “AI”. Generally speaking, people are referring to what is generally called “generative AI” in these cases.

You know all those insane tools in the adobe suite that can do stuff like literally erase a person from a photo or weather a surface or even select only the object you want to delete with a single click of the mouse? Those are, varying levels, of the same underlying algorithms behind “AI” content creation. Hell, most of the good plugins for IDEs for handling stuff like docstring or unit test stub generation are in a similar boat.

By and large, people don’t have major issues with those. Some of the training data gets really messy but they are a fundamental part of most creative workflows and can be argued as being comparable to using a reference book when drawing human anatomy and so forth.

The issue comes when you take that a dozen steps farther and have “generative AI”. Rather than take an existing photo and remove the ex you hope dies in a fire, you just say “hey grok. Make a photo of me on the streets of Osaka by myself. And undress a child while you’re at it”. Rather than create a docstring or unit test stub you just have Cursor write an app for you based off a prompt. And so forth.

At which point it stops being a case of someone using the same reference material to draw a superhero and more that guy who just traces porn for Marvel every month.

And… much like someone who can’t draw their way out of a paper bag, you see the same with generative AI use in content creation. Generative AI is generally great at replacing entry level employees. It can’t replace a skilled senior creative. And if you are wondering how you get people the experience they need to hit that tier… you get it.

But that leads to the other problem. If you are someone who is cutting costs left and right to increase profits and realized you can replace 60% of your staff with a subscription to openai? How long until you decide that if you just lower your standards a bit you can replace 80% instead?