EAGE 2021
Conference paper



Artificial Intelligence (AI) has been successfully adopted in many industries in recent years. The results are encouraging, with AI being able to reduce costs and improve performance in different applications, sometimes outperforming its human counterparts. However, most of current models and technologies are still restricted to specific tasks and cannot be easily adapted to different contexts without a significant effort. Such an ability is especially important for knowledge-intensive tasks such as seismic interpretation, which is heavily dependent on the interpreter's experience and tacit knowledge. Moreover, this dependency makes it challenging for oil companies to deal with interpretation biases and knowledge loss when, for instance, seismic interpreters leave the company. To tackle these pressing issues, we propose a system that sheds light on the transition from Narrow AI to the so-called Broad AI, in which we combine powerful machine learning models with an efficient knowledge representation and a symbiotic human-AI interface. It is a first step towards broad AI, and the results obtained so far have shown the system's ability to better manage the corporate knowledge, reduce bias and improve seismic interpretation quality and time requirements.