Surpassing human capabilitiesPhoto: Pixabay

Surpassing human capabilities

A recent global machine learning competition resulted in new algorithms that can predict lithology from well logs as good as, or even better than humans.

More than 400 teams consisting of close to 2000 individuals from around the globe attended last year’s FORCE Machine predicted Lithofacies competition.

The background and the motivation for the competition, according to the organizers, is the fact that the current revolution in data science and machine learning is fuelled by open data sets. They refer to open data sets as a democratization of tools and real time access of knowledge and innovation.

At the upcoming digital DIGEX conference, Peter Bormann (ConocoPhilips) will present the results from the competition. The top 10 algorithms were all very close in terms of their ability to determine lithofacies from well logs alone, and the machine predictions were on par or even surpassed human interpretations.

A few examples of a “human” lithology interpretation next to a “machine-predicted” lithology. Also presented here. Source.

Further improvements to the algorithms, datasets and the way the competition was run are currently being researched by a global community of geodatascientists.

All the data, the submitted algorithms and the interpretation of the results can be accessed here under an open source CC-BY-4.0 license.