AI as a Creative Partner (and Forecaster)
A Look Back at Three Generative AI Projects — Part 1 of 3
First, some context
Over the past two years, I’ve had the chance to design and deliver three different projects that use AI in one form or another.
All these projects were made on invitations by C2 Montreal, an international business and creativity conference that also offers participants the opportunity to attend Labs: experiences designed by artists that explore a facet of their practice.
Across these three projects, a theme has emerged: an exploration of what human-AI collaboration could look like. Each of these projects has been, in its own way, a good teacher on this subject. I wanted to take the time to reflect on this and share what I learned while designing and producing them.
I created them with the intention of providing participants with a first-hand experience of interacting with (or thinking about) AI systems and how they impact us — a way for a non-tech audience to poke at the tools with specific goals in mind.
They are also, I think, somehow representative of some aspects of AI adoption and public literacy at the time of their creation (ranging from 2022 to 2024). I think of them as time capsules, or ways to look back at the incredible rate of breakthroughs that became mainstream over such a short time span.
With that in mind, let’s dive in!
A Better World
In September 2022, A Better World was the project it all started with.
The premise of the experience goes like this: a group of 15 people who don’t know each other have 40 minutes to come up with an idea that can change the world (for the better, hopefully). The idea can either be something very concrete (a device, an app, some form of product) or abstract (a law, a process, etc.).
Why use AI?
The main idea behind the project was to see if and how AI could help a diverse group (made up of people who don’t necessarily identify as being ‘creative’) come up with ideas they could be proud of.
Over the months leading up to that invitation, I was seeing people from various backgrounds using GPT-2 to generate creative outputs so I was curious to give it a try. At the time, I also stumbled upon the concept of Augmented Intelligence, the notion that generative AI systems can run in parallel with human creative endeavours and support it rather than replace it.
The initial challenge (you have 40 minutes to come up with world-changing ideas) was absurd by design: if you have such a short time, why not use all the help you can get?
A creative cornucopia
At the beginning of the experience, people were provided with hundreds of ideas that were generated by the AI (GPT-3 at the time). These ideas were curated by me, refining ways to prompt the system to generate proposals in various categories, inspired by the UN Sustainable Development Goals — which provided a base framework for ideation.
The process of generating those ideas took quite a bit of time to get familiar with the model and, through trial and error, get it to produce propositions I’d find interesting (often because they’d lead to more questions and be conducive to conversations).
The approach to prompting I ended up using was a sort of cascading one: starting from one of the UN goals, I would prompt the model to provide a series of strategies. Then, after having cherry-picked a collection of strategies, I would prompt it again to provide a list of concrete inventions that would embody said strategies.
My objective at this point was to strike a balance between quantity and diversity: there had to be enough unique ideas to convey a sense of abundance and appeal to people from diverse backgrounds.
The ideas were then made accessible in the form of long strips of paper — a medium that felt approachable, inviting, and tactile.
People could use these ideas as a starting point for conversations. Each participant got to share which ideas they picked and why they chose them.
This conversation helped people align. By listening to what each of them valued, they got to know one another and were able to land on a collective objective they wanted their upcoming idea to tackle.
A sounding board
At this point, the AI also played the role of a counselor: if participants had a hard time reaching consensus on which objective might be the most worthy of pursuing (should we prioritize access to healthcare or ending gun violence?), they could turn to the system for an opinion.
Once they reached a consensus on which area to focus, participants entered a (human-)guided brainstorm. During this phase of idea generation, three behaviours seemed to emerge. Participants were either:
- Remixing: Using the machine-generated idea they had initially picked as a start point and tweaking it to some extent to fit the collective objective.
- Combining: Merging their or other people’s ideas into something new.
- Starting from scratch: Coming up with ideas unseen so far, while sometimes using pre-made ideas as examples of opportunities or challenges.
While they brainstormed, participants still had access to the counseling support of the system: they could input their idea and get a pros/cons analysis. In this mode, the system helped our aspiring innovators identify potential shortcomings — shortcomings they might have missed in the heat of their eureka moments.
AI played the role of some kind of all-knowing entity, a test drive for ideas, useful in that it brought a virtually infinite number of external perspectives on the creative output of the group.
A forecaster
At the end of the process, the group could submit their final idea to the system and get a prediction about how it might unfold over the next 50 years.
The point here was not to be accurate (who can predict the future anyway?), but rather to invite people to think about the long-term consequences of their idea — the good, the bad, and the ugly ones.
From the feedback I gathered from participants after the experience, the consensus was positive. People left impressed by their own ability, as a group, to come up with what could be a pitch for a start-up in such a short time span. The human-likeness of the AI’s comments left a strong impression on them.
They didn’t seem (yet) to fear the mind-boggling rate at which this type of technology would enter their daily lives…
📡 Stay tuned for Part 2!
