Is AI my work colleague or my replacement?

Edit: This week's newsletter is missing several images, and I am unable to upload them. For a full version of this edition, feel free to visit: https://marekkowal.substack.com/p/generative-curiosity.

Yesterday, I spoke at a small gathering organised by Futures Collaborative Brisbane, where we explored the use of artificial intelligence in futures thinking (also known as corporate foresight). I have had a bit of experience in this space. During the session, I shared some of my thoughts about using Generative AI in futures thinking with participants. This week’s post is a newsletter-friendly summary of my 10-minute talk.

Artificial Intelligence (AI) is sparking conversations in boardrooms, living rooms and even restrooms. Businesses “appoint” AI algorithms as CEOs (which, oddly, always take the form of a female-humanoid robot). Every week, we hear about an AI algorithm beating humans at yet another skill. And so it’s no wonder that many of us ask a question: is AI destined to be a collaborative partner in our work, or is it edging toward replacing us altogether? The answer, unsurprisingly, is not that simple.

Generative AI, the subset of AI that’s capable of crafting narratives, designing graphics, and composing melodies, is at the heart of this debate. This isn’t the AI of yesteryear (or yestergear) that streamlined data analysis or automated mundane tasks. No, Generative AI is entirely different: its sole purpose is to mimic what a typical human-generated content would look (or sound, or even feel) like. At their heart core, these algorithms are trying to be… like us (no, they don’t have a will or desires, but they do have an optimisation function that rewards them for being more like us).

A New Ally in Creative Thinking

Things get interesting in the field of creativity (I wrote about it earlier). Consider scenario planning, a task that organisational strategists perform, requiring quite a lot of creativity. Scenario planning involves a deep understanding of the organisation, its environment and any potential trends that might impact it in future. And then it requires the ability to imagine plausible futures. Until recently, the notion of a computer contributing anything of value to such a nuanced task would have been dismissed. But now, Generative AI is flipping the script.

Here’s an example. In 2022, before ChatGPT became available, my team used OpenAI’s APIs to GPT-3 to generate futures based on the data we collected during various workshopping activities. It was a bit geeky, but it worked. Importantly, it saved us a lot of time—we could generate a scenario in less than 10 seconds—about a thousand times faster. You can get a glimpse of that in the text in the bottom right quarter of the screenshot below. This is a massive improvement compared to possibly a couple of hours when doing this manually.

Such future scenarios are used to elicit insights in workshops. Before building our tool, we used four scenarios per workshop. With the new tool, we suddenly had the option to have four scenarios per participant and hundreds of scenarios per workshop! It’s really powerful (and also challenging to manage, but that’s for another post).

This example is a year old. In today’s pace of progress, this is eternity. Admittedly, you’re probably quite used to AI-generated text. Still, I think the example remains quite impressive, given its particular domain application!

Should we be worried? The fear that human-led scenario planning may become obsolete is understandable but also misplaced. AI is impressive, but it’s not an autonomous force—it thrives under human direction. These futures would be useless without us taking them to a facilitated discussion and deriving insights based on them.

Generative Curiosity

This is where I propose we introduce a new term: Generative Curiosity.

Historically, curiosity has been the spark of progress. Curiosity is a crucial trait for human evolution. This innate trait has propelled us from the savannahs to the stars. ChatGPT wrote the previous sentence when giving me feedback on this article. I decided to include it here when I saw how cheesy it was.

Design Curiosity came next, embodied in design thinking. Through a systematic approach to innovation, design thinking helped us be curious, even when that curiosity wasn’t innate. Design thinking, and thus design curiosity, thrives in many organisations for good reasons.

Retrospective curiosity followed. It is a strategy that leverages existing data to unearth new insights, akin to analysing years of sales data to forecast market trends. Imagine being curious about what has already happened. Retrospective curiosity is a mindset in which you imagine and experiment and then assume it has already happened. All you need is to find it in your past data, and then you’ll be able to see the experiment’s outcome. Sounds strange? Plenty of social media networks experiment just like that—by looking into the past. Here’s a research paper that describes one such experiment.

What emerges now, and what you saw in the screenshot above, is Generative Curiosity. This form of curiosity leverages the power of Generative AI to explore futures at scale (quantity) and the aspects of the future that some of us wouldn’t even think about (quality). Algorithms are the new partners to ask curious questions about the future. Some AI researchers define such algorithmic curiosity as “the error in an agent’s ability to predict the consequence of its own actions”, which pretty much means that an agent is curious when it doesn’t know what might happen next (and, presumably, decides to act upon it).

And using Generative AI in futures thinking is pretty much it: trying to get our algorithms to a point where they cross the domain of the known into the unknown. In most cases, “hallucinations” of Generative AI are a problem. In futures thinking, they are something we desire!

In most cases, “hallucinations” of Generative AI are a problem. In futures thinking, they are something we desire!

Your Team of Algorithmic Futurists

If your job is to imagine futures… imagine the following. With generative curiosity, you could harness large language models as your very own digital minions. You could run a collective of chatbots, each equipped with distinct preferences and perspectives, ready to churn out scenarios by the dozen. And then, you could have yet another group of digital minions that could collect all these scenarios, group them, and prepare for the next stages of futures thinking. Given the potential this creates, it’s hard not to think about them as something more than just tools; they’re, in some algorithmic way, partners in ideation.

Your algorithmic futurists!