I was asked to write a blog post about AI In the oil and gas industry. I will try to give you my approach about Artificial Intelligence in this industry where everyone can conduct a working proof of concept but struggling to get to industrialization stage.
First, I would like to warn everyone attending management presentations where you hear: "This is complex, I don’t exactly know how to do it, but let’s put some artificial intelligence models, it will save me.” Run away!! This is the rabbit hole; it will never happen.
Before I clarify how to react to it, let’s define artificial intelligence so that we all talk about the same thing in this article. The definition of artificial intelligence that I will use in this article is:" A computational system, which is capable of observing sensing its environment, reason on it, learn from it, and surface a decision on which we act on.".
Ok, now that we have a definition, when someone comes and would like you to solve their problems, ask these 2 questions, and map them on a graph. "Do you know the answer of your problem? Is your problem complex?". If it falls into the quadrant "Complex, and I don't know, or I can't validate the answer"…. Run away! You will not be able to solve the problem in a reasonable amount of time and it will not be adopted. Where you want to embrace your customers is where they know how to resolve their problems, and it is fairly simple. If it is a repetitive painful task, you win even more. Why? Because your user will be able to validate your decisions, you will save them time to make these decisions and they will be able to dedicate that saved time on the more complex problems and simplify them or dedicate time on problems they don't fully understand , explore them and get into the understand zone.
Now that you have a good chance to be able to help your customers, ask yourself where do they fit in the digital world? If you want to increase the success of implementations into production, map them on the digitalization maturity scale. They need to have mastered the workflow of their problem, they need to have digitalize their workflow, and they need to have explored their data via analytics. Only then, AI should be tackled. If the first three steps are not there, you will get stuck into the Proof of Concept, the research, the "we can do it, but we can't integrate." because you will have no place to insert that AI solution into and act automatically on the decision that will be surfaced.
If you apply these last 2 concepts, you are on a good track, but you can increase your success rate even more. For that, you will see that if you control your environment, and reduce the variance, you will simplify and accelerate your AI solution. An example if implementing a deep-learning approach from a moving camera feed with different angles, distances, etc. compared to a fixed staged camera set that is there to perform one task. You will simplify your learning, you will reduce the parameters, accelerate the solution and increase your success rate.
I have now exposed the three key approaches to increase your changes to have success in AI enabled projects. If you reach that stage and want to go the extra mile, then don't stay narrow minded and too focused with your data scientist cells. Once you have successfully delivered your AI solution to the business, don't think that you are done…. Follow up! In order to reach the nirvana and have your AI product deployed in production, you will have to accompany the business to adopt and trust your solution, teach them that the model will have to live and be improved. AI is not a "implement and forget".
With that short and to the point blog, I hope you will be able to accelerate your AI ambition and implementation. For a future post we could talk about Ethical AI, Explainable AI, and Frugal AI. This is an exciting time, and exciting world where we should rename Artificial Intelligence into Augmented Intelligence.
Author information: A Geek and a Photographer who is still shooting regular films in medium format. Passionate about solving problems, his moto is “Make it simple; initiate purposeful impacts”. Currently Director of the Artificial Intelligence Lab at Schlumberger in France, his first rule is common sense!