Black Box Problem

The Black Box Problem in artificial intelligence, notably in the world of ML (Machine Learning), refers to the opacity related to the internal workings and decision-making process of AI models. These complex models—such as deep learning networks—often generate results that are difficult to interpret, which can be problematic in settings where the transparency of the decision-making process is crucial, like healthcare, finance, and legal. This poses a significant challenge in terms of trust and accountability, as it's often pivotal for developers, end-users or stakeholders to know why and how the AI arrived at a particular decision.

The Black Box Problem became more prevalent with the rise of ML in the 21st century—it has grown in prominence with the expansion and complexity of AI. When the AI systems were simpler, their decision-making process was easier to grasp, but with growing complexity, the model's inner workings became less clear. The term came into popular use with the rise of deep learning algorithms in the 2010s.

While it's difficult to assign the development of this concept to specific individuals, researchers and organizations across the globe have contributed to the understanding of this problem. It's now a central issue in not only computer science but also in areas like ethics and law. Recent years have seen increased attempts to solve the black-box problem using techniques such as LIME (Local Interpretable Model-agnostic Explanations) and SHAP (SHapley Additive exPlanation), aiming to create 'Explainable AI'.