The “uncanny valley” refers to the eerie feeling humans feel when they encounter a not-quite-human robot that’s meant to resemble us. Usually the term spurs thoughts of physical, humanoid robots, but we’re here to tell you it applies to the generative AI models that have gained traction recently. That’s right, the uncanny valley has become a mainstream problem.
Oh, come on. That really surprises you?
Try using DALL-E 2 or Midjourney to produce a modern painting of a human. Chances are it will get key features wrong, and not in an artsy, Picasso way. We’re talking missing fingers, or too many fingers. We’re talking unnatural facial asymmetry and menacing eyes that pierce into your soul. Why are those eyes looking at you like that? Is the AI angry with you for exploiting it for cheap labor? No, the likelier reason these images come across as creepy is that we’re not quite sure what to make of them. They’re almost good, but usually not yet “human quality.”
The same goes for writing. In fact, we tried using AI to write this piece. The model didn’t produce anything erroneous—even though generative AI chatbots are known to occasionally lie. In fact, it actually wrote something that was quite intelligent. The problem is it wasn’t engaging enough. It was missing the wit, color, humor, and snappy prose. That’s a huge problem, especially for content penned by artificial intelligence public relations professionals that’s meant to be captivating enough to catch the attention of busy reporters.
That being said, AI can offer great support in your PR efforts. The key is not to overdo it.
Titles aren’t AI’s strong suit
PR pros write many titles, ranging from press release titles to subject lines for pitch emails and events submissions. In many ways, they’re the hardest part of any writing task, for two reasons: 1. Your title is the first impression you make on your reader, so it better be good and 2. Getting your point across with fewer words is always more challenging. As such, titles require a high degree of human ingenuity that AI simply can’t match yet.
That’s true for tweets and other short-form social posts too. An artificial intelligence public relations pro might dabble in ChatGPT for titles, but it’s probably better to bounce ideas off AI than to rely on the titles it writes for you.
AI can help with ideation and outline-building
The uncanny valley in AI writing doesn’t have the same “creepy” factor as it does in AI art. It’s not as though a shiver will run down the reporter’s spine when she reads a pitch email penned by an AI. In fact, she might not even catch that ChatGPT wrote your entire press release.
She won’t know AI wrote your content, but she will know it’s not up to par. It won’t be engaging enough. It will be too wordy, rigid, and, yes, robotic. Artificial intelligence public relations professionals should absolutely leverage AI to bounce ideas around, build outlines, and even rewrite paragraphs. But relying on ChatGPT to actually do the writing for you is a bad idea. That’s especially true considering it has no idea what happened in the news yesterday, a week ago, or even a year ago.
As for the images…
PR pros usually aren’t responsible for the images a reporter will run along with the story they write about our client. Nevertheless, there are instances in which we have to pick between a logo, a stock photo, and using an AI-art generator to create our own image. The latter is often the fun option, and it can also be the most effective.
The key, however, is to nail the prompt and ensure you pick something that actually works with the story. Just because something looks kind of cool for AI art, doesn’t mean it should be published in a story about your client’s funding round. Always ask yourself—would I pick this photo if I had found it in Pixabay, too?
That’s one kind of AI bias—actually being overly excited about the fact that this is AI-made. The other kind is being inherently biased against AI art. Yes, it can produce some uncanny-valley images, but it can also paint some really impressive stuff. For people with an anti-AI bias, ask yourself whether you’d feel as turned off by the image if it appeared in a high-end art gallery under the name of a famous painter. Are you sure it’s really that bad?
As these models continue to advance, artificial intelligence public relations pros will have to adapt, but chances are we’re not getting out of the uncanny valley anytime soon. Learn to use these models, accepting their limitations, and you will reap the maximum rewards.
