AI big data news

BP’s algorithms replace pumpers. BP seismic guru havers between computational physics and ‘possibly overhyped’ machine learning. US NSF 2019 update on AI R&D and standards. Matlab’s predictive maintenance toolbox.

It is customary for those advocating the application of artificial intelligence (and not just in oil and gas) to conclude a presentation with reassuring words along the lines of “of course AI will not replace you, it will ‘free you up’ to concentrate on more productive work”. This is not how things worked out for BP’s Wyoming-based pumpers whose jobs, according to an article in Forbes, are being ‘taken over by algorithms’. BP has deployed San Francisco-based start-up Kelvin’s AI to monitor real-time data streams and optimize production from its Wamsutter oilfield. The result is that BP needs 40% fewer workers to keep its natural gas flowing in Wyoming. What’s more “field techs are now getting trained in Linux and Python.”

In a ‘Seismic Sound-off’ podcast from the Society of Exploration Geophysicists BP’s John Etgen opined briefly on the yin and yang* of conventional computational physics (CP) based seismic and novel machine learning (ML) approaches. With some understatement, Etgen said that there is a chance that the promise of ML is overhyped. ‘I'm not going to predict that the CP approach is dead. But it is drawing itself into a corner where every new advance requires an order of magnitude increase in compute power and makes the fundamentals of the inverse problem much more complicated to solve. We are making life harder for ourselves. But I'm not sure the ML stuff is going to deliver everything we want either. There’s good work to be done to figure out what that optimum pathway actually is. Other data intensive domains that are not math/physics based, where you cannot "prove" what the truth is, are moving towards ML. We scientists and engineers are not moving as fast in that direction as some others are. We’ll see how it turns out. I don't know.’

* See our Watson and the weather update elsewhere in this issue for more on this.

The Computer and Information Science and Engineering division of the US National Science Foundation has published a short introduction to the US National AI R&D Strategic Plan. AI began back in 1956, when a small group of computer scientists and mathematicians met on Dartmouth’s campus and first coined the AI term. Investment in AI is now a national priority, with notably a 2019 White House Executive Order on ‘maintaining American AI leadership’ and the American AI Initiative. The 2019 update to the National AI R&D strategic plan describes work done to develop AI-relevant standards including P1872-2015 (Standard Ontologies for Robotics and Automation and the standardization program of ISO/IEC JTC 1 SC 42 on AI. Those looking to benefit from the NSF’s largesse need to visit the awards searches page.

For a quick introduction to predictive maintenance, download the free Matlab handbook, a 17 page walk-through describing how to identify condition indicators and discriminate between healthy and unhealthy machine states and how to develop, train and deploy ML models in production. The software is available as a trial.

Click here to comment on this article

Click here to view this article in context on a desktop

© Oil IT Journal - all rights reserved.