On 21 May 2026, I spoke at the 20th anniversary Infoshare in Gdańsk. On the Marketing & Sales stage, I had 20 minutes for my talk "Is creativity repeatable? Images, systems, responsibility".
I went through three phases of creative work with AI: generating ideas, copying and transforming, and making a selection. The most important takeaways:
- A model will not generate the same image or video twice. Even with highly repetitive prompts, we get something different, because that is how these models 'think'.
- Creativity is repeatable as a process, not as a result. From a single shot, I create several variants in contextual models, and once I build a complete workflow, I just need to swap the products.
- AI cannot count. It once generated a sofa with six panels, and another time with five, so someone has to check it.
- Models can generate characters from famous films on their own, even with completely safe prompts. The person who releases such graphics or video into the world is responsible for them, as the companies creating the models often distance themselves from this.
- AI will not do the final stage, the selection, for us. We will remain the guardians of taste and quality here for a long time to come.
You can find the talk description on the Infoshare website.
Photo: Infoshare