Cold and snowy outside, but warm and enthusiastic inside the Geological Society at Burlington House for the Janet Watson’18 meeting dedicated to the impact of Big Data in Geoscience. As soon as the event announcement was published on the web few months ago, it was fully subscribed in the following hours. This illustrates pretty well the current pick of expectations for Artificial Intelligence in geoscience.
Some technologies such as the 3D seismic, the integrated earth model, the virtual reality have fascinated the O&G subsurface community in the past decades. But I have never seen this level of hysterical enthusiasm for new tools as the SVM, CNN and Deep NN which have not yet produced so much in geoscience.
Listening the presentations (full program is here) and discussing with the attendees in London this week, I think it may be explained by at least 3 reasons:
At first it may be because O&G engineers had got the feeling their industry was at the leading edge of the technology. But they discovered recently than retailers, virtual travel agencies, entertainment providers implement now much important IT stuff for storage and computing than they do. So, our first behavior is to adapt the same tools than the GAFA and other NATU, praying their magic will move us to the top position again.
Another reason of the current enthusiasm is the feeling of freedom provided by the new AI tools: They break the silos and and come with datalakes where you can fish as you want everywhere you want, they also allow you to program interactively with your data, almost intuitively (Jupyter Notebooks), and finally, they open a world of interactions with academics, startups, social media instead of limiting the discussions with few specialized software providers as most of us were stuck until today.
This enthusiasm comes also from the AI tools capacity to deliver rapidly some first interesting results: It is possible to re-use a DeepCNN developed to detect dogs and cats on pictures, to detect faults and salt domes in few weeks. Do you need to develop something more sophisticated? There is a chance over two that you can be inspired by a GitHub something (library, API, Apps or even interactive tutorial). Do you need to access more subsurface data to train your models? Ask the North Sea NDRs (National Data Repository) for data: The OGA in UK, the NPD/Diskos in Norway and the TNO/DINO in the Netherlands are now in a push mode for data. The OGA, the CDA and the BGS have explained this week their open data strategy and their action to support AI initiatives to create values from their data assets. Do you need more open subsurface data: Visit and access the numerous virtual outcrops detailed this week at the Society of Geology like the SAFARI project
Finally, it could be said that the current enthusiasm is not only for a new technos but much more to a new way of working and finding solutions to subsurface issues.
But several lecturers have also alert us about the danger of this current level of expectation. In a interesting paper, John Turmond from Statoil remembered us the Dunning-Kruger theory.
We don’t learn anything and cannot predict anything just remaining at the top of the Expectation mountain. To gain experience, it is necessary to practice and if possible to fail, to feel desperate in the “valley of despair” before to find the avenue for some real progresses. We got some insight this week of what could be the future avenues of progresses in geoscience. The Geological Society should publish special issue containing the papers which have supported the presentations. No doubt that some of the works presented this week will deliver some real progresses in the domain of geoscience in months or years. If you want to know which of them have the more potential of solving efficiently some subsurface issues, the best is to go to practice, to experiment, to engage with academics, startupers or opendata providers, to put your hands in oil as usual! I am sure that most of the people I met this week have started this AI journey or will do it shortly and consider, as I do, they are currently living the more intense and more interesting period of their geoscience career, the AI revolution!