Why Is AI So Bad

Why is AI so bad? Or, rather, why is bad AI so bad.

Sure, it’s almost miraculous that you can conjure up virtually any text, image or video in seconds, but AI’s uncanny valley is more like a Grand Canyon. From fake humans with their weird goblin fingerclaws in b-roll to overbulleted, slop text, AI has become a byword for the degradation of the web and content in general.

And bad AI is also a reputational risk for companies.

Coca-Cola’s recent holiday ad was bad AI’s big debut – a nauseating mix of smiling humanoids, beverage truck tires that seem to hover over the snow and quick cuts to mask the visual incoherence of the sequences. Some PR and marketing professionals called the ad “soulless” while the beverage maker stood its ground. Researchers found that consumers didn’t know or didn’t care about the AI since, after all, the holiday ads have been CGI for years. 

Bad AI wins!

To be clear; bad AI is not good

But Coca-Cola is sui generis. It can afford to ride a wave of controversy, perhaps to show investors its embrace of cost-efficient technologies for its eight-or-nine-figure marketing budget. 

But don’t be fooled; bad AI is not good for companies. It makes them look sloppy. It makes them look like they don’t pay attention to details. It makes them look they don’t think humans can tell the difference. 

I’m not an AI skeptic. I got into natural language processing years ago because of its promise as an analytical tool. The thought of having 1,000 robot interns poring over reams of text and Moneyballing PR was tantalizing. But marketers and PR professionals are mostly interested in AI’s generative capability, which creates nearly infinite variations on content for pennies. 

I say nearly infinite because the one variation AI seems to struggle with is the one that’s actually good. 

How do you know you’re looking at bad AI? The corporate stock photography looks fine at a glance but, up close, you see malformed faces. The b-roll in videos looks glossy but is impossible to place; marquees and signs have mangled text; humans and, even animals, seem to glow and have nothing behind the eyes. 

“Count the fingers” is the new “follow the money”.

New rules for AI usage

Inevitably, AI will improve. But the path is going to be uneven. Until then, we’re recommending that our clients exercise care and judgement when using AI in PR or marketing and follow a few rules:

  • On more permanent marketing platforms, like the company’s website, don’t use AI imagery. Use stock photography, or even better, real images of your technology, your facilities and your people.
  • For one-off or special-purpose projects, use a mix of AI stock and real b-roll, particularly when depicting humans. Look closely at the finished products. 
  • Use AI as a copy editor for already created text, not the other way around. Text and copy should be run through AI only once it’s already written. 

Humans can prevent you from being embarrassed by bad AI.