You Can’t Build a Future You Can’t See

Stories can make possible AI futures tangible, not by predicting what will happen, but by helping us see possibilities we might otherwise miss.

Jules Verne did not build a submarine. He did not travel to the Moon. And he was certainly not very good at predicting exactly how either would eventually happen.

But decades before many technologies became reality, millions of people had already experienced versions of them in their imagination.

That may be one of the things we are missing in the discussion about AI.

We have forecasts, benchmarks and increasingly impressive demonstrations of what AI can do. What we have surprisingly little of are convincing pictures of what our work might look like if we took those capabilities seriously.

The lasting value of Verne’s stories is that they allowed readers to experience a possible future before it existed. Machines became part of adventures, institutions and human decisions. Their consequences felt real.

We could use stories in the same way for AI.

Technical descriptions tell us that future AI systems might maintain long-term memory, coordinate other agents, review engineering evidence or participate in complex decisions. But such descriptions leave important questions unanswered.

What would it feel like to work with a system that remembers every decision made during a 20-year aircraft programme? What happens when it challenges the chief engineer using evidence that no individual in the room can fully verify? Would engineers become more capable, more dependent, or both?

A story could place the reader inside that design review.

The AI recalls why a structural margin was accepted twelve years earlier. The original engineers have left the programme. The current team has always treated the decision as settled. A young engineer notices that one of the underlying assumptions is no longer valid.

The story gives us a setting in which to consider authority, memory, responsibility and professional courage.

What stories add

A forecast asks which future is likely. A story asks what a possible future would mean for the people living in it.

AI capabilities are changing rapidly and unevenly. Tasks that appear beyond AI today may become routine within a few years, while other expected advances may prove much harder. How organisations, regulators and people respond creates further uncertainty.

Stories allow us to explore that uncertainty without pretending to resolve it. One might present AI as a trusted engineering colleague. Another could examine whether dependence on it weakens human expertise. The value lies in considering different possibilities before one of them becomes reality.

We tend to see the future through the present

Our attention is selective. Once an idea becomes relevant to us, we start noticing connections and possibilities that previously passed unnoticed. Imagination can work in a similar way. A future that we can picture becomes easier to recognise in fragments around us.

Without such pictures, we tend to imagine new technology inside the structures we already know.

We see a requirements process and ask how AI could make it faster. We see a report and ask how AI could draft it. We see a design review and ask which parts AI could automate.

Those are reasonable questions. But they assume that the future will retain the processes, documents, interfaces and organisational structures of the present.

Once we can imagine a more fundamentally different way of working, other questions become visible. Why does this report exist? Would this review have the same form if every participant had access to an AI with complete programme knowledge? Which processes are technically necessary, and which exist because coordinating large groups of humans is difficult?

Stories give us room to explore those questions before we have to implement anything.

A laboratory for engineering questions

For aerospace engineers, such stories could explore questions that technical demonstrations rarely capture:

• Who is responsible for a design decision developed through hundreds of interactions between engineers and AI systems?
• How can a certification authority assess reasoning that changes as the system learns?
• What expertise must humans retain when machines perform most routine analysis?
• Will complete digital memory improve safety, or make organisations reluctant to reconsider earlier decisions?

A useful story would remain close enough to engineering reality to be credible. It might introduce one major change, such as an AI system able to follow an entire programme, and then examine the consequences. The technology does not need to be described in detail. The human dilemma does.

I have written one such story for my own department. It explores a possible future of our work as a way of making that future concrete enough to discuss, criticise and improve.

Before an organisation can prepare for a future, its people must be able to imagine it. Stories can make abstract questions concrete and give engineers, managers and regulators a shared situation to discuss.

This is why AI may need its Jules Verne. We need technically informed stories that are willing to follow emerging capabilities beyond the boundaries of today’s organisations and processes. A good story will have served its purpose if it helps us notice a possibility, question an assumption or make a better decision today.

You cannot build a future you cannot see. Stories can help us see more than one.

References

• MIT News, “Jules Verne, desperado?” — on Verne’s use of contemporary science and technology in fiction
• Michael Burnam-Fink, “Creating narrative scenarios: Science fiction prototyping at Emerge,” Futures (2015)
• “The narrative aspect of scenario building — How story telling may give people a memory of the future,” Springer

AI helped to research and write this article.