© 2016 Agile Data Decisions LLC.                                                                                                                              Proudly created with wix.com

  • YouTube Social  Icon
Site map
More information

Follow us on the social media

Download our flyer

Download a recent "The leading Edge" paper about iQC

Please reload


She won't feel alone ...


Please reload

On release

EAGE 2017 - WS01: A revolution ?


Illustration from Matt Hall keynote available here

GitHub for this mix of face image and seismic here


For the first time, EAGE organized a workshop about data sciences during its main annual event in Paris this year. We can measure the way done by the geoscientists for data science adoption in less than 2 or 3: our industry is going fast!

Remember, two or three years ago, very few data sciences papers were presented in conventions such as the EAGE, the SEG, the SPE or OTC and at this time, most of them were high level and philosophical papers about the 3 V and how our industry should embrace the Big Data (sorry for the authors of several exceptions to this rules!).

Even a year ago, most of the paper were marketing papers overselling the AI magic to geoscientists who are fan of physics and know that "correlation is not reason". This scarred me since the possibility of disappearing in an ultra deep through of disillusionment is real after a too high pick of expectation!

The number of EAGE workshop 01 papers was a first good news. The variety of real business cases presented illustrated pretty well that data science, AI and ML are now adopted by all the branches of geoscience from from seismic monitoring to seismic compression and storage going through remote sensing interpretation, log QC or interpretation.

The first keynote by Michel Lutz, TOTAL group Data Officer, illustrated pretty well how a large IOC is mobilizing the right technical and human resources to take advantage of the recently renewed AI techniques.

Teledetection using satellite or aerial images by ONERA and TOTAL


I appreciated a presentation and a poster done by A. Smith of Lunchelan and M. Wildig from Ovation Data who explained an experiment of seismic storage on Hadoop which didn't deliver the expected results. I consider we never publish enough on the things which does't work as well as expected. Sharing this type of information is more valuable that any speech on something which goes well.

In addition to the speeches, a good point is that the organizers have let some space and time for the posters. Poster are a good way of interaction btw the authors and the attendees for a real exchange of experience.  I was obviously attracted by the  poster named "The Unstructured Data Challenge in E&P" by Phoebe McMellon from Elsevier. She works on subjects very similar to ours with geofacets, we have to keep in touch!


So, could we say now that everything is perfect in a perfect world perfectly described and analyzed by machine learning?


The second keynote presented by Matt Hall from AgileScientific and available here was a perfect demonstration of the challenge we still have to face. According to Matt, we are just starting a "revolution" in geoscience. AI cannot just be considered as a new tool in the geoscientist tool box in addition to the physical equations we use to model the subsurface. AI is efficient today because it comes with Big Data and Big Data causes or must cause a real change of attitude for all of us.

These changes of attitude have been perfectly listed by the attendees in the word cloud below thanks to mentimeter, a presentation tool used by Matt which allows the audience to interact with the speaker:

Mentimeter word cloud captured by Matt during his keynote


As you can see in the middle of the graphic is "OPEN DATA" and not so far away is "OPEN SOURCE" and on my opinion, "OPEN INNOVTION" could have been added.

This illustrated that a lot of the attendees were ML practitioners and know that the learning models we build are efficient only if based on large amount of various data and experience sharing. OpenData, OpenSource and more generally experience sharing using new communication protocols are for them the natural ways to improve their solution and convince the geo-community AI works. This explains that most of the question to speakers were about accessing their data, API, source code ... 

If the oil companies and main contractors go really in the direction of a more open environment in the next decade, Matt would have been right: We are at the beginning of a revolution.




PS1: I like this post about the EAGE workshop described above. Jesper Dramsch is a young and plenty of talents Geophysicist ᑎ Coder ᑎ Writer!


PS2: Matt view on the event is here . I guess I missed plenty of other posts on the same topic....


PS3: You missed the workshop but want to know more about the presentations: Just go to EarthDoc. Our abstract is accessible by clicking on the image below:




Share on Facebook
Share on Twitter
Please reload

Please reload