Nature study discovers data drift and its impact on AI performance

Recent research published in Nature highlights how AI models trained on unreliable or unverified data can produce inaccurate predictions, undermining their effectiveness. While this study focuses on medical imaging, the issue of data drift—the gradual decline in AI performance due to poor-quality data—applies just as much to energy management.

AI might have a place in enhancing energy monitoring and optimisation. Still, if AI models rely on flawed, outdated, or AI-generated data, errors can accumulate – leading to miscalculations in energy usage patterns, inefficient resource allocation, or even increased consumption. For AI to truly support sustainable energy outcomes, businesses must ensure their energy data is reliable, validated, and traceable. Without this foundation, AI remains a powerful assistant but not a decision-maker when it comes to managing energy in the built environment.