Technical Assurance in Four Important Pieces:
- AI with Domain Expertise
- AI as a Decision Support System
- AI that Uses Representative Data Sets
- AI that Delivers with a Domain Expert
Forbes recently honored Petro.ai as one of the 15 most innovative companies in AI right now and highlighted the “seismic shifts” these companies are creating in their industries. Petro.ai’s revolutionary approach to deep learning is already making dramatic inroads in moving shale to an economically competitive place in developing the energy our world needs. Technical assurance is important with any AI system and the multi-million-dollar decisions that O&G face every day need to be made with confidence and accuracy. That confidence and accuracy is built into the AI of Petro.ai.
AI with Domain Expertise
“There are four major considerations that provide technical assurances within Petro.ai,” Dr. Derek Ruths, Chief Data Scientist of Petro.ai begins. "One is that we are domain experts. There’s a real danger with people thinking an AI system works just because they’ve trained it on data. What makes Petro.ai work is a lot of actual domain know-how, like Dr. Mark Zoback our Science Advisor, and the deep bench of engineers at Petro.ai, all designed into the system. They know this science and they know the basins involved in shale. Petro.ai is a full integration of a tremendous in depth understanding of the rock and the full well life cycle.
AI as a Decision Support System
“At Petro.ai we understand that AI itself is really a decision support system. Not a decision maker. And that means we’re empowering engineers with an incredibly resourceful and accurate tool that will help them ask the right questions and help them look at their wells in a different way. Ultimately, the decision sits with them. Our expectation is that when people use our AI they go in as critical thinkers. That’s part of what’s important about the product, it’s used by knowledgeable, competent invested clients. Clients who have the background and expertise to make these decisions.
AI that Uses Representative Data Sets
“At Petro.ai we embrace what I think are the most important design principles of AI and foremost where that’s concerned is representative data sets. We don’t try to make the problem easier by choosing data that’s going to be simpler for the system to learn. We choose data that represents what is in the field. That means there are a lot of edge cases included in the way the AI is trained. When we say that it gets 96% accuracy that’s estimated using blind tests on real data as opposed to training and testing on data that we’ve seen before.
AI that Delivers with a Domain Expert
“Another important aspect of the product is that the delivery involves a domain expert on our side. When we set up the tool in the client’s environment, when we upload their data, when we orient them to the dashboard once we’ve prepared the analysis, a Petro.ai expert is on the call. He walks through an assessment of what he’s seen. That’s also another place for the client to ask a question such as, is this curve really here or is it just tracking one particular well?Clients need to be critically thinking about the results that come back.
Always Technically Assured and Continuously Improved
“Petro.ai follows a continuous improvement regimen where when more data becomes available, we add that data. As we deploy and do projects and work with clients, whenever we see something that looks a little unusual, we always investigate. There’s always this continuous improvement mentality. And that’s really important. You can be assured that Petro.ai is trained to provide the highest levels of accuracy possible.