It's all about the information you put in

Before generative AI, there was Photoshop. If that sentence makes no sense, bear with me a moment.

Generative AI models have the power to seemingly make something out of nothing. That’s wild! Years ago, graphic designers had to tell managers that no, we can’t just magic something out of thin air in Photoshop. Coders had to deal with leadership not understanding it takes time to write code. Writers had to actually write (or hire ghostwriters). Now, with the magic of generative AI, any person can wave their wand and make their heart’s desire appear. We’ve surely surpassed our limits, and can create information from nothing… right?

Well, no. I don’t think that’s the complete picture.

Photoshop wasn’t (and isn’t) the only image editing tool out there, but it’s famous. At one point, it was the most popular image editor by far. It had basic editing functions, complex editing functions, brushes, layers, tools, and filters. Even if you weren’t great at using Photoshop, chances are you used its filters. My uneducated guess is that filters are the main thing that made Photoshop famous, given how they seemingly added something to images out of nothing but a few clicks of a mouse. In some cases, you could make a completely new image from scratch.

Reality is a bit more complex; Photoshop filters were never magic. They were the result of knowledge, engineering, and testing, all encoded in software form for individuals to call upon. Put another way, filters are a form of learned information that were packaged in an easily usable form. There’s a slight parallel to generative AI models, which are the result of lots of knowledge, engineering, and testing, all encoded in software form to be called upon. Photoshop filters may not work at the same scale as AI, but neither are truly nothing.

That doesn’t mean those managers were right in asking graphic designers to just “Photoshop it.” All the filters in the world couldn’t give you the ability to make something good without training, effort, or intent. The best graphic designers studied and practiced their craft, and it showed. Casual users like me could… add cool borders and make some neat color changes. You wouldn’t hire a casual Photoshop user.

However, we are more accepting of casual generative AI use. The scale of training that goes into generative AI makes AI a whole lot of something. It’s much easier to create something of usable quality with generative AI than with Photoshop filters alone. When a tool like Claude can generate a pretty website or a great spreadsheet template in a few words, of course we’re impressed! Or at least, we’re impressed until the cracks start to show.

If you use generative AI without intent or context, you may start to notice problems. Conversations become inconsistent; documents look strange; websites have weird bugs; images look same-y. You keep trying to fix the problem, maybe not realizing that the AI model has anchored itself to a concept it can’t forget. In rare cases, the model starts going against your orders. You certainly created something, but not something good.

The people who are more successful with generative AI are not using it casually. They use AI with intent, thought, planning; they read articles; they listen, practice, adapt, write, and keep up-to-date with domain knowledge. They carefully select new information that the AI model should know to better steer it. In other words, they study and practice their craft, developing the necessary skills to build something good. As a technical writer, I can confidently say that generative AI only creates decent documentation when the user leverages domain knowledge and personal skills.

So yes, before we had Generative AI, there was Photoshop. Both can “create something from nothing” by leveraging engineering to generate data. In the same vein, both need skill to truly create something of quality. The big difference is that Photoshop’s worse filters could never reach the level of a basic image made by ChatGPT.

We may not like that more people now ask us for something out of nothing, but we can at least take comfort in knowing that when we get good results, they came from real effort.