Are we copilots or just passengers on ‘AI flights’?

Human agency in the age of AI

A few days ago, while boarding a flight, I glanced into the cockpit. It was a maze of screens, dials, levers, and buttons—even on the ceiling! The confident smiles of the pilot and copilot reassured me, and we landed smoothly. In my seat pocket, I found an ad for Microsoft’s Copilot, an AI tool promising to solve business and societal problems. It made me wonder: are we copilots on this “AI flight,” participating in navigation, or are we just passengers, content to fasten our seatbelts?

Unfortunately, the reality is that we are mostly passengers. We understand very little about how AI functions, how it uses our data, or even how reliable its outputs are. But it doesn’t have to be this way.

Why does it matter to understand AI?

AI is not just another tool. In our flying analogy, AI isn’t merely the plane that takes us from one city to another—it can also decide why and where “we should fly.” AI has become an integral part of our daily lives, shaping personal decisions and global trends alike. It powers driverless cars, smart home devices, and social media algorithms that shape our opinions.

This is why understanding AI is crucial. It’s about being more than passive recipients of technology’s impact; it’s about actively shaping our reality and future.

Is it difficult to grasp AI functionality?

It is not. At its core, AI is built on two simple concepts: patterns and probability. These are the building blocks of everything from basic machine learning to advanced models. We don’t need to be experts to understand that AI’s probabilistic nature can lead to hallucinations or misleading answers. Grasping the basics can help us be more discerning and less likely to be misled by AI-generated information.

We must not perceive AI as an untouchable, mysterious technology that turns us into passive passengers simply “fastening our seatbelts.”

How do we become real AI copilots?

First, we need to understand how AI generates its answers—from the powerful hardware that runs it to the data we feed into it and the algorithms that transform that data into user-friendly outputs. Knowing the full path from input to output reveals AI’s capabilities and limitations.

Second, we must be aware that AI answers are influenced by the ‘weights’ assigned to data used by AI. For example, the future UN AI system must give the highest weight to the UN Charter as a key document, supporting organisations’ fundamental values. 

Third, we need to ensure the quality of the data used in AI systems. The old IT adage, “garbage in, garbage out,” also applies to AI. AI’s computational power means nothing if it processes flawed or biased data.

Fourth, we must reconsider our approach to bias in AI. Aiming for completely bias-free AI is unrealistic and potentially dangerous, as it assumes a single truth. While we should condemn illegal or harmful biases, we should also recognise that some biases reflect human diversity and values. AI transparency is crucial for understanding whose perspectives and values a technology reflects. With diversity in AI development teams, transparency is key to mitigating risks while acknowledging that certain biases are an integral part of human experience and decision-making.

Next steps?

To use and govern AI effectively, we must understand its core components—computation, data, and algorithms—and the risks they pose. Computation involves the infrastructure to run powerful AI systems, while data and algorithms determine what these systems generate as answers, images, and other artifacts.

By learning the basics, we can move from passive passengers to active copilots, taking control of our lives in this AI-driven world. Being a copilot means more than just understanding technology—it means participating in discussions about how AI should be developed, regulated, and used. It means advocating for ethical practices, supporting diversity, and ensuring that AI serves the common good. Only then can we truly navigate—not just ride—this AI journey.