AI and the Danger of Good Enough

In late 1999, if you were heading to a Y2K party and wanted high-quality photos of the event, you had two choices: You could either spend thousands of dollars on high-end camera equipment and learn how to use it (often in paid classes) or you could hire someone who had done all those things for you.

Cell phone cameras were brand new (and terrible) and, though standalone digital cameras existed, they were of low quality and still very expense. If you wanted something higher quality than a point and shoot camera, you needed both skills and equipment, whether your own or someone else’s.

But, times changed. Digital cameras got better and better. Now, any reasonably recent smartphone has a better camera in it than even existed 20 years ago, and the technology powering those cameras make it easy for anyone to take decent photos.

Can you still get better photos with high-end specialized equipment and expertise? Absolutely. But for most circumstances, a cell phone camera is seen as good enough. The quality is more than adequate, and the convenience of sharing from an internet-connected device is tough to beat.

When you head to a Y2K23 party this year, you’ll likely see people taking photos with their cell phones rather than hiring photographers or breaking out specialized equipment.

That kind of photography wasn’t stung by a better product. Professional photos are, almost always, objectively higher quality. Instead, it was beaten out by a free and convenient alternative that was good enough. 

That, in turn, is the real danger with AI images and AI writing. Not that they will beat what real human artists and writers can do, but that they will be good enough for many applications.

That reality already seems to be here, at least in some areas, and it’s going to call for a reevaluation of what humans really do best.

Stupid, But Adequate

In a recent article for The Atlantic, Ian Bogost highlights some of the limitations of what ChatGPT can and cannot do.

The biggest limitation that he highlights is that AI can’t really understand human language. It’s attempting to respond to prompts by placing words in a particular order to deceive the user that it’s understanding.

To highlight this, he attempted to get ChatGPT to create poems and the bot struggled, misunderstanding both the meter of its work and what makes a work of a particular genre.

However, even with those limitations, in Bogost’s testing the bot managed to write an acceptable 5-paragraph student essay on itself and the only criticism that he had of it was that it was “consistently uninteresting as prose.”

That is undeniably true, but much of the writing we humans undertake is meant to be functional and isn’t meant to be interesting prose.

You can see this where AI is already being applied. In 2017, five years ago, the Washington Post revealed that an AI bot named Heliograph had produced some 850 articles for the paper. However, those stories were for things such as local high school football games and financial reports. They were all short, formulaic and not worth sending a human reporter to.

At the time, the stories had generated more than 500,000 clicks. In short, they were ridiculously successful, delivering information that was important and sought after, but also not requiring human input.

So, even if we agree that current AIs are dumb, they are still adequate for many of the tasks we would normally give to human authors. 

The Sky is Not Falling

All that being said, I don’t think the sky is falling when it comes to AI either. While I agree that academia has some challenges ahead of it, I don’t think the essay is dead nor do I believe bots will replace human creativity.

What I am saying is that, just because AI doesn’t write as well or make art that is as good as what humans can do, doesn’t mean it won’t have an impact. We’ve already seen AI art win art competitions and, as with Heliograph, seen AI writers produce content for an unaware audience.

The immediate end game of AI isn’t to beat humans at their own game, but to tackle the tasks where “good enough” is more than adequate. This includes anything that is rigid, formulaic and has a set of easy-to-understand rules that the AI can follow.

An AI can interpret a box score or an election result, but it won’t be writing feature pieces of athletes or analyzing the impacts of an election. It may try to and even create prose that seems to fulfill that need, but it won’t hold up to much scrutiny.

However, that takes us back to the core problem. There are times and places where filling the space is more important than filling it well. Returning to the example of camera phones, it’s more important that those images exist and are shared than they be of perfect quality or say anything deeper than surface level.

Much of the reading we do and images we look at are not intended to be art, analysis or anything deeper than exchanging basic information. That is a space that AI has already proven itself capable in.

Bottom Line

To be clear, the camera phone didn’t get rid of the photographer, professional or amateur. It did force a reckoning about when such a photographer was needed and how to best use those talents.

Photographers have struggled, but their occupation is projected by the U.S. Bureau of Labor Statistics to grow some 9 percent over the next decade. There is a similar projection for writers, albeit with only 4 percent growth

In the end, AI is a new and advancing technology. It will have impacts on both creative professions and academia. However, what those impacts are will only be clear in hindsight.

However, those impacts aren’t about AI doing more work or better work than humans. It’s about doing work that is good enough for many situations.

With that said, AI is already there, at least in many applications. Whether it’s writing short articles or creating stock imagery, Ai can do an adequate job when the rules are clear and it “understands” the assignment.

As such, it’s not a matter of if, but how AI will operate alongside human creators.

It may not be the most comforting though at the end of 2022, but it is certainly a reality we’re facing as we head into 2023.