In a thought-provoking analysis, Tony McGrail, Solutions Director for Asset Management & Monitoring Technology at Doble Engineering Company, sheds light on the limitations of Artificial Intelligence and Machine Learning (AI/ML) systems in data analysis.

McGrail shares his experience of using an AI language model to interpret anomalous condition monitoring data and emphasizes the significance of human expertise in understanding contextual information. He highlights that while AI/ML systems can provide valuable insights, they struggle to generate new knowledge and often require human interpretation for complex scenarios.

McGrail points out that AI/ML systems build upon existing data and lack the ability to extrapolate effectively from unseen or unfamiliar situations. The article showcases an actual case study where an AI language model failed to provide a conclusive answer when faced with anomalous behavior in monitoring data for a power transformer. Ultimately, McGrail emphasizes the need for subject matter experts who possess domain knowledge and experience to interpret data accurately, especially in critical areas like power infrastructure.

The article further discusses the challenges and potential biases associated with AI/ML systems. McGrail highlights examples where AI tools exhibited flaws such as misclassifying images, showing biases based on training data, or lacking transparency in decision-making. He stresses that while AI/ML systems can be beneficial in certain areas, the expertise of human analysts remains essential for addressing complex and unique cases, which may require thinking "outside the box". Read the full article in the May edition of our magazine.

This article was originally published in the May 2023 issue of the Advanced Transformer Testing and Technologies magazine.

View Magazine