Adaptability to Change
What is it?
AI systems are often trained on large datasets collected at a specific point in time. This means that, in many ways, AI reflects the world as it was then — not necessarily as it is today. In a constantly changing world, this can become a problem if AI is not continuously updated, adjusted, and challenged.
This applies to attitudes, language, demographics, culture, and societal conditions. An AI model may perform well in the months following its training — but over time, it risks becoming inaccurate, biased, or outright misleading, because its understanding does not evolve with reality. Therefore, adaptability — both technical and organizational — is a crucial part of responsible AI usage.
Examples:
A clear example can be seen with the image generator Midjourney, where prompts for an illustration of a "doctor" often produce images of older white men. This is because the model was trained on global image data that reflects outdated and stereotypical perceptions — not necessarily the current reality in, for example, Danish hospitals, where today there are more women than men and considerable ethnic diversity.
Another example is AI translation tools, which often misinterpret expressions when language use has evolved over time. A classic case is the Danish word "bjørnetjeneste," which traditionally means "an unwanted favor" — but which many young people today interpret as something positive. If an AI model is not updated with current language usage, it can misinterpret or distort messages, even when the translation is grammatically correct.
What to consider?
Consider where and how data obsolescence might affect your use of AI. If you use pre-trained models, find out when they were last updated — and be aware that they may contain outdated norms, obsolete facts, or changed contexts.
Establish procedures for the continuous updating, evaluation, and critical use of AI systems. This is particularly important if the AI is used for communication, classification, translation, or content generation that needs to be current and culturally relevant.
Adaptability to change is not just about technology — it is also about culture and awareness: being ready to intervene when reality shifts.