Shy AI Creatives

There’s a concept in modern-day political polling of “shy Trump voters” that intend to vote for Trump but don’t want to mention it. So this means his polling is always a bit lower than his actual support levels. On the other hand, we have people who love Trump, and proclaim it loudly. And I think people using chatGPT in a creative context might have a similar issue. Some people are loudly proclaiming how great it is, but I think far, far more people are quietly using it and not interested in stirring up a backlash.

I seem to run into 10 skeptics for every person openly admitting they’re finding real value from AI. Some of my friends have written it off. Some of my coworkers have aggressively called us out for using AI-generated art, which led to a lengthy and anguished response from me, the company’s head of design, about where and when we will use AI.

(Tangent: My argument boils down to the idea that I’ve been in the field for nearly 25 years, and I can remember my teams paying for a photo or illustration twice. Most of the other time we’ve used free stock photography, just like how must people raid images.google.com for their Powerpoint presentation rather than setting up a film shoot. This, combined with the idea that of course you shouldn’t purposely scrape an artist’s work to make rip-offs of their art, combined into our stance that we were going to use AI art, despite understanding the backlash)

But I’m here to tell you that creatives in your life are absolutely using AI, and they just don’t want to tell you about it. It’s similar to when Twitter came out and no one could understand that it was more than “what you had for lunch.” Or that blogs could exist about anything other than “navel gazing.” I know I’m going to lose my audience when I say it here, but it’s the same reflexive mindset that is surprised (and maybe a little concerned) to see a woman pilot. It is the same thing, even though you’ll want to argue with me. It’s all one simple little word, one that causes the most damage when we think we’re safe from it. That word is ignorance. If you don’t know, you don’t know, and you pretend you do.

Let’s clear up some misconceptions. First, the idea that it’s just mindlessly putting one word after the other is simultaneously true and completely missing the point. Yes, that’s how the technology works, but if that’s your mental model you’re not going to understand how significant the technology is.

If I ask Google “Do you have suggestions for writing E2E tests for my web app?” then I will get a list of ten articles, as you’d expect. If I ask the same question of chatGPT, it starts with some popular tools, then general steps, then writes a sample test for me (with install instructions), then shares best practices, then asks if I have any specific questions it can help me with. See the response here. One word after the other? Sure. Just like how Romeo and Juliet is just a love story. This is missing the point.

Second, hallucinations, the word used to say AI makes things up. In the popular imagination people seem to think chatGPT is spewing garbage, then humans are copy/pasting responses without looking them over, and then submitting entire wrong answers to the judge overseeing their case or the teacher teaching their class. And that has happened, to be sure. But there’s a much more interesting story lurking beneath that one, which is that there’s a universe of value you can get from things that have no single right answer.

For example, let’s say I’m writing jokes with a comedy writing partner. We’re bouncing ideas off each other and I ask my friend “I feel like there’s a joke in here somewhere where we talk about systematic racism in Hollywood in a way that really makes the audience think. Maybe we can bring in Mel Gibson? Maybe there’s a Will Smith angle?” and from there we’d riff. Well, I proposed the same thing to chatGPT and it did a fantastic job. See the response here.

In this response, chatGPT outlines five themes to explore and then gives a few proposed jokes to help illustrate the point. Are the jokes good? Would I use them verbatim? Well, no. But are they going in the right direction? Could that 10 second query provide me with more forward progress than a one hour brainstorm? Absolutely yes. chatGPT isn’t for objectively right answers. It works really well to bounce ideas and iterate on concepts, so the criticism that “sometime it gets facts wrong” is obscuring the real story.

Third, the idea that all creatives are against AI is just wrong. They are much more concerned about being canceled or ignored and right now there’s not a lot of benefit to admitting you’re using a tool people think is evil and bad. Saying you use AI makes you sound immoral, lazy, and maybe an NFT scam bro. So there’s not a lot of advantage to mentioning it, so people think the whole creative world is against it. We’re not, and I can prove it. It’ll just take a few years for people to look back and admit that they too were once a shy AI creative, back in the early days of 2023. But we’re not there yet.