Our Art(ist)ific(ial) Intelligence
AI can do a lot, does a lot, and is everywhere today—so naturally it’s also part of our agency. One thing always remains the same, though: We humans like to get a little crazy, experiment, mess around, in short: we’re creative. AIs are not creatives, but analysts—at a certain point, they start going in circles with their “thinking”. However, in the interplay of artificial and artistic energies, great projects can emerge—and sometimes quite a bit of nonsense. We’re sharing some of our experiences here.
AI Fails
Give the AI assistant a bit more freedom, and suddenly it becomes clear that it’s missing something—namely, common sense. The Lindau island suddenly shrinks to just two buildings, and the appetizing schnitzel enjoyment turns into an unappetizing gorging fest.
Man eating schnitzel
4-man bobsled
Lindau Island
Icons on the topic of water treatment
And the fact that a four-person bobsled normally has a somewhat different seating arrangement seems to slip the AI’s mind… well, but at least it creates beautifully designed, well-thought-out, and easily understandable icons upon request. (Warning: sarcasm!)
Our
AI THURSDAY.
Recognizing opportunities and risks, discussing breakthroughs and fails, sharing hopes and fears—that, or at least something like it, is the idea behind our biweekly AI meetings. At the pace new models, features, and applications are coming to market, keeping a close eye on the latest developments is absolutely necessary, whether you're a graphic designer, copywriter, or account manager. That's why we get together every other Thursday, talk about interesting new tools, compare our experiences working with AI, and collect our findings on a miro board accessible to everyone. This serves not only to increase efficiency in our daily work, but also generally to familiarize ourselves, get hands-on, reduce apprehension, or in other words: to settle into an AI-shaped future. It's sometimes funny, sometimes eerie—but always highly interesting.
AI Text Fails
Let a chatbot decide what sounds good? You can forget that in most cases. Especially when finding new product names—like here for a water filter—it often completely lacks the feel for which creations are both appealing and pronounceable—not to mention the meaningfulness of some suggestions…
ChatGPT “Apology”
But correcting texts, ChatGPT can do quite well—most of the time, anyway: When asked to check a website for grammar and spelling, the chatbot responded by calculating away for FIVE DAYS and constantly explaining upon inquiry that it was working with great enthusiasm. When it was finally asked to present an interim status of its work so far, it delivered… a completely empty Excel document. Well, at least it was able to explain its failure in detail afterward and apologize profusely.
