Brainstorming with GPT-3 — Our Wow-Journey!

We are four students studying interaction design at the University of applied Design in Schwäbisch Gmünd, Germany.

In our course “Invention Design” we are exploring emerging technologies and matching meaningful use cases that bring additional value to a specific group of users. Our team decided to work on AI-assisted brainstorming for designers. Lucky us, we got acces to an AI called GPT-3 from OpenAI. GPT-3 is a language model that uses machine learning to generate natural text.

“Brainstorm means using the brain to storm a creative problem.”

— Alex F. Osborn

Brainstorming is all about creating many different ideas regarding a creative problem. They can range in different directions and don’t have to be bound to a specific frame. Creative problem-solving is characterized by two processes — variation and selection: At first you collect a wide range of ideas, which don’t get valued immediately. The next step is to narrow the solution space down with specific methods and to evaluate which solution is the most suitable.

GPT-3 presents itself as an entity, that doesn’t replace designers and their work, but more likely assists them in different parts of the design process (e.g. ideation) and accelerates these. The AI is capable of creating an associative solution space in creative brainstorming, while imitating and communicating methods used by designers at the same time.

“How might we“-questions (HMWs) are used to phrase problem statements. They often reveal a problem space while already including specific insights from the previous research. These questions work really well when given to GPT-3 for brainstorming. During our semester we collected a bunch of HMWs together with fellow students and professors and gave them to GPT-3. As the generated outputs were way better than we could ever imagine, we would like to share them with You:

How can we teach people to change their own habits more effortlessly?

How might we help employees stay productive and healthy when working from home?

How can we help blind people find their way more independently in unfamiliar environments?

After the comparison between the AI-generated output and conventional „human-made“ brainstorming and discussions with professors and fellow students we figured out, that all involved went through a sort of „learning-experience“. We consider this experience being of enduring relevance for designers. For us, it consists of the following aspects:

Experiencing this Wow-Effect, we reset the goal of our project. Eventually the question: „How can we share our Wow-Effect with other designers?“ arose.

To answer this question we manifested our insights and learnings, and had many conversations with our colleagues on how creative, generated output might affect the work of designers. These are the things we came up with:

1. One thing is certain, namely that we will interact and work with machines more frequently in the future than we do now. In the future, creative output will come not only from people, but also from machines.
2. In the future, design skills and other subject-specific skills will not only be tied to a specific job description, but also to general purpose AIs.
   That is why designers must find new areas in which they can specialize without having to compete with AIs in the long term.
3. Thus, in the near future, designers will probably rather take on a selective and curating role in large parts, especially with regard to variation and selection in the design process.
4. In conclusion, it is up to all designers to decide how much artificial intelligence will soon influence our everyday work, since we as designers are in a position to regularly question our tools and shape them ourselves.
5. Looking on the state of AIs now, designers will not be abolished or displaced, but merely supported in creative solution findings. Therefore, we can speak of a kind of “positive” disruption which we should not be afraid of but use it as a tool that we can shape.