When Knucklepuck conceived of the AI Olympics, our goal was simple: create a fun, engaging way to showcase AI tools and create a practical use case for the importance of human input.
By asking the team to use ChatGPT and DALL-E to generate a logo for our upcoming company retreat in Denver, we sparked creativity but also experienced firsthand the limitations of generative AI and the consequences of being overly reliant on these tools.
We certainly got some exciting designs, though after several attempted iterations within ChatGPT we realized that it was time to take things “offline” to get them across the finish line.
Taking the Handoff
Rather than prompting ChatGPT to make further changes to the image, I asked for instructions that I could hand off to a graphic designer to make the changes manually.
ChatGPT provided an initial outline of specific design change requests that I documented for our creative vendor. This allowed me to significantly reduce turnaround time vs. drafting something from scratch and also tapped into ChatGPT insights for enhancement opportunities.
The next iteration helped bring a lot of these design changes into focus:
But the logo still had issues. The 2024 was poorly framed, the spacing of DENVER, CO looked odd, and KNUCKLEPUCK didn’t pop the way I wanted.
No more ChatGPT! Now back to the old school approach of annotating specific screenshots with explicit instructions:
By the end of only 2 rounds of revisions requiring less than 3hrs of manual design work, I finally felt this was the polished product that deserves to be printed on our retreat t-shirts!
The Moral of the Story
ChatGPT and DALL-E made me 10x the graphic designer I was before. But this still was not enough to produce a true finished product.
As we learn how to incorporate AI tools into our workflows, it’s critical that we’re conscious of the law of diminishing returns. While there are tremendous gains to be made with breakthrough technology, as we continue to tap into the same tools the value-add of each output decreases. In this example, as we got closer to our “final product” ChatGPT was less effective at making substantial improvements. Eventually, the tool was no longer useful and we needed to pivot toward manual tactics.
This experience taught us that while AI can be powerful, human input is still an essential part of the creative process. ChatGPT and DALL-E provide foundational elements and guidance but ultimately creative work still relies heavily on human intuition to produce quality graphics that showcase our brand, vision, and expertise.
