Machine Learning: Is it Magic or Hard Work?
Ridvan Akkurt (Schlumberger)
Machine Learning (ML) has become a key tool to analyze large datasets, detecting previously unforeseen patterns or extracting unexpected insights in high-dimensional space. ML workflows can process large volumes of data efficiently with less bias and subjectivity introduced by humans. Unfortunately, the present level of excitement in the industry, fueled by many publications that highlight only the successful case histories, has created a sense of magic where ML (or Artificial Intelligence) is considered the solution to many difficult and challenging problems. As in the case of any other emerging technology, the power of ML, if unleashed without proper controls and practices in-place, especially without the understanding of the uncertainties in the answers, will lead to erroneous results, adversely affecting the acceptance of the new technologies in the long run. The objective of this talk is to identify the requirements and the critical components that are needed to build robust ML systems; and illustrate their impact and effectiveness by sharing examples from real-life projects. The industry can benefit from the new technologies only by being careful, cautious and realistic; and with the understanding that to get to the magic of ML, one must be prepared to do the hard work.
To get to the magic of Machine Learning, we need to be prepared to do the hard work.
Ridvan Akkurt is a Petrophysics Advisor in the Schlumberger Artificial Intelligence and Analytics Group, based in Denver, USA. He was previously Research Director at Schlumberger-Doll Research in Boston, heading the Geoscience Department. Ridvan began his career in 1983 as a Schlumberger wireline field engineer in Africa, then worked for GSI, Shell, NUMAR, NMR+, and Saudi Aramco; in international and domestic assignments. He has a BS degree in electrical engineering from the Massachusetts Institute of Technology, Cambridge, USA, and a PhD degree in geophysics from the Colorado School of Mines in Golden, USA. Ridvan has many publications and 24 US patents, has taught industrial courses, served as a Distinguished Lecturer for SPE and SPWLA.