What do the two sciences need to know about each other to deliver practical results? On the image above is a wavelet displayed to scale next to a wave-shaped building. Appreciation of scale is one of the skills critical in oil and gas AI projects. On the other side, to be a successful petroleum geoscientist in the post-COVID requires a set of very specific new skills in addition to the understanding of geologic concepts. These new skills are not normally associated with geoscience but are fast becoming critical for geoscientists to operate in an E&P environment.
To be a successful petroleum geoscientist in the post-COVID requires a set of a very specific new skills in addition to the understanding of geologic concepts. These new skills are not normally associated with geoscience but are fast becoming critical for geoscientists to operate in an E&P environment.
In a lot of ways, the COVID pandemic provided a perfect opportunity for oil and gas companies to make some long-delayed changes to their structure and how they operate. Changes to both of these affect the job of the geoscientist, as the structural changes mean there are fewer available entry-level geoscience jobs to fill and the operational changes mean a heavier reliance on applications that integrate artificial intelligence and machine learning into geoscience workflows. To do more with less, companies realize they must leverage these new technologies; the issue they are facing now is finding people who know enough about both topics to be effective at applying AI technology to answering geoscience questions.
Herein we argue lies the opportunity - those geoscientists that can interface effectively with new technologies are going to be in high demand. Knowledge about structure, reservoir and depositional systems needs to be supplemented now with knowledge of data science principals and machine learning architecture. The geoscientist needs to understand the algorithms deployed in data science and the implications to the results of interpretation of geoscience data.
It is important to understand the underlying mathematical principles of machine learning because most programmers that design such systems do not have any geoscience background and are unaware if the network output violates geoscientific principles. For example, the concept of geologic superposition is something that all geoscientists are aware of and the principles of which are rigidly adhered to. It is not however a widely-understood or contemplated topic among...well...anyone else. Particularly not computer scientists and software architects.
Presented here is a short, non-comprehensive list of the top four skills petroleum geoscientists of all experience levels should consider building and/or refining. For experienced geoscientists currently working in the industry, the list of skills to acquire or augment should help with designing the roadmap for future geoscience operations. For hopeful entrants into the industry this list provides some key suggestions about how to steer their current educational curricula and early industry years to make sure they are able to hit the ground running and stay relevant in their newly acquired oil and gas positions.
Geoscientists are now – by necessity - data scientists too. The tools that are becoming available to geoscientists in the oil and gas industry require large but clean and well-organized data sets for input. Only geoscientists know what the data means and how to gauge its quality, so they are instrumental in deciding how the input information looks and - based on that - how reliable the output is.
To that end, a short list of things that are critical to a successful petroleum geoscience career in the next decade:
1. Learn the basics of AI. There is some very easily understood mathematics that explain how neural and - at a larger scale - deep learning networks function and “learn”. Become familiar with the fundamental equations and terminology such as “feed-forward”, “back-propagation”, “loss”, “accuracy”, and “learning rate”.
2. Read a book/take a course about human psychology. One of the best ways to understand the underlaying principles of AI architecture is to understand how the human brain learns and develops. These systems have been designed to mimic human learning functions, so understanding what they are modeled after makes understanding their architecture and use easier.
3. Take the extra math classes. This one applies mostly to the “hopeful entrant” category but could apply to everyone. It’s worth it to be able to understand some of the more complicated mathematical concepts behind machine learning -the minimum required level of mathematics to attain an MSc in Geology is insufficient to engage with machine learning architects in an effective way.
4. Familiarize yourself with cutting-edge software applications. Applications such as Bluware’s InteractivAI™ represent the type of significant advancements that have recently been made in geoscience machine learning. Become an expert user and early adopter of one or more of these technologies so you can explain at a high level how they work and how to interpret the output information.
5. Understand and accept that the job has changed. There was a point in the not-too-distant past that hard-core geoscientists (paleontologists, sedimentologists, geophysicists, hard-rock petrologists, etc.) had their secure place in oil and gas. With very few exceptions, this is no longer the case. The harsh reality is that some specializations are simply no longer relevant to finding or producing hydrocarbons. The geoscientist of the future will be: 1)an informed data custodian; 2) an organizer of meaningful populations of data;3) the interpreter of the output information*; and 4) Perhaps the most important - a decision maker.
(*the key here being to formulate an interpretation at the end, not at the beginning.)
Ironically enough, the very fact that neural networks are built to mimic human neural behavior also makes them vulnerable to noise or biases baked into datasets. Overload them with conflicting or irrelevant information and they get confused. For example, collecting seismic data from all sedimentary basins in the world to automatically interpret all possible fault configurations in a new basin does not always produce good results. Quite frankly, such an endeavor should be expected to fail simply because of the wide range of ways faults are/were/will be represented on seismic data. The historical nature of oil and gas data adds an additional dimension of variable data quality, which is another layer of complexity that requires a geoscientist to make sense of. Machine learning networks - specifically the things they learn - are different from basin to basin just as the geology is different. The characteristics of one cannot be expected or assumed to fit the reality of another.
There exists a healthy but delicate balance between human experience and machine learning; the key to succeeding as a geoscientist in the future is to successfully bridge the technical gap between what experienced geoscientists know (the wisdom)and what machine learning algorithms can learn (the knowledge).