By Geoffrey Cann
New sophisticated artificial intelligence tools are coming to oil and gas. There’s plenty of upside, but lots of worries too.
A few years ago, a fried who had read one of my articles happened upon a suspiciously similar article posted on LinkedIn published by another person and forwarded it to me. On inspection, the article had been quite blatantly copied from my blog series, word for word. The plagiarist worked as a strategy advisor in the strategy office of the world’s largest oil and gas producer at the time, which was awkward to say the least.
Out of politeness, I connected up with the chap on LinkedIn, pointed out that the article was obviously a cut-and-paste job of my work, and if there was some reasonable explanation. He immediately ghosted me, left the article online and under his name, and so the gloves came off. Read on for the resolution.
There could well be many other instances of such plagiarism of my work, and who would even know? It was only discovered by accident. For those of us who earn a living from composing words to communicate ideas, theft of our intellectual property has been one of our biggest worries (second only to the worry that no one will read our work in the first instance).
ChatGPT Enters The Fray
Language processing has been one of those things that humans have claimed to be uniquely of our species. Sure, other creatures are known to communicate with each other using sound—whales sing, dolphins chirp, and dogs bark—but only humans have created symbols to abstractly represent language, to allow language to be taught, to codify its usage, and to reduce the potential for misinterpretation.
Using automation to translate the sound of the human voice to text, and to translate text into human voice has been with us for many years. These capabilities are now bundled for free with your phone and built into search engines. They solve for accessibility barriers experienced by those who are unable to work with the typical screen and keyboard interface. An entirely new class of product, the home assistant, is based entirely on these capabilities. They work within languages and across languages (translation and interpretation).
In hindsight it’s no surprise at all that a more embracing and capable automation engine for original text generation would surface. A cool new product, ChatGPT, which was launched just 90 days ago, is already challenging quite fundamentally the world of language processing, by directly creating content, which has exclusively been the human job. No more reliance on the human voice to form the words, or the fingers on the keyboard to type the letters.
These artificial intelligence tools are experiencing extraordinary growth. Anecdotally, to run ChatGPT is costing a $1b/month in compute costs, in just 90 days since its launch.
How ChatGPT Will Eat A Chunk of Oil and Gas
Imagine a bot with unlimited patience, unlimited but flawed technical depth on a range of topics, continuously learning, fluent in multiple languages, globally consistent, universally available, supremely self confident, frequently wrong, and massively parallel. Imagine if that bot relied instead on much internal corporate resources, and not exclusively on the flawed information sources of the wild internet.
How might such a content machine impact the oil and gas sector?
Manning question/response services — many industry websites feature chat bots, but these have been quite simple in the past, providing at best pre-crafted information. Imagine the ability to have deeper unlimited conversations with unlimited participants on the fly.
Enhancing equipment interfaces — the interface between much equipment in oil and gas places huge demands on the human operator, requiring deep training, service manuals and instructions. Imagine replacing the manual dumb interface with an intuitive, intelligent, competent, voice activated interface that lets less experienced operators perform like pros.
Writing instructions and guides — the industry devotes considerable resources to crafting instructions and guidance on equipment and procedures, for safety, reliability and performance reasons. Imagine a bot with the ability to crank out these resources at scale.
Composing narration — the industry is very text heavy, with 300+ page annual reports, billions of emails, regular information circulars, compliance reporting, applications and permits, training course content, legal contracts and agreements, incident reports, and business reports. Much of the composition is mundane and susceptible, whole or in part, to automated generation.
Crafting social media — the industry has generally refrained from engaging too vigorously with social media (Twitter, YouTube copy, Instagram, LinkedIn posts). As one exec put it, zero engagement in social spheres starves the social media world of its oxygen. These tools will be able to generate an unlimited level of quality outbound content.
Enhancing internal search — one of the compelling digital use cases in oil and gas is the ability to instantly search historical data (reports, analysis) to find just the right bit of useful insight. Automate this for a huge engineering time saving.
Engaging talent — I always struggled to write a compelling end of year performance review. Imagine all of your employees turning to these tools to boost the quality of their year end discussions. What happens to recruiting when job ads receive a flood of highly customized near perfect applications? Performance reviews, improvement plans, official correspondence, and policies will soon be machine generated.
Coding — a shortage of coders will give way to the ability to instantly generate code for whatever application needs that arise.
Building engineering content — two years ago, the big engineering houses such as Worley and Wood were deploying automation engines to build the mundane P&IDs used throughout the industry. These AI tools will soon enhance procurement specifications, requirements documents, and bring a level of automated engineering to every engineer.
That’s just a snippet of what these tools will be able to do.
For every good use case these tools present, they also introduce new complexity to the already demanding challenges facing oil and gas.
Cyber attacks — if these tools work to write code for good, then they can also write code for bad, and at scale. The industry already struggles with the onslaught of cyber activity, which will likely intensify because of the war in Ukraine.
Flooding the zone — to the same degree that these tools can be put to use generating high quality social media content, they can also be used to generate high quality and false narratives about the industry. Scam email and correspondence will increase. Instead of receiving five thousand identical form letters individually signed in some kind of petition from the public, companies can expect to receive hundreds of thousands of individually and beautifully written correspondence on key topics.
Bot on bot mayhem — what happens when two language generators interact with each other? There’s a high likelihood that the chat room does not involve an actual human engaging with the oil company bot, but another bot. Could a chat between two bots generate offensive results, and could those results be used to harm a company’s interest?
Busted supervision — the speed, cost, and productivity gains from these tools are clear, but the risks and liabilities associated with the content generated by such tools isn’t settled. Who owns the liability for an inaccurate result being generated by a bot, and relied upon by an unsuspecting user? Intense supervision of the output of these tools may be required given their track record so far.
Impacts on the human workforce
Once these tools take over more and more communcations, what is to become of the human worker whose existence to date has been linked their technical skill and their ability to communicate that knowhow to others? These tools will inevitably generate false results to some query or request, and in oil and gas, trust in these tools can evaporate very quickly.
Some digital innovations, such as 3D printing or blockchain, have been set to the side by many in oil and gas. However, these new language bots are different. They hold too much potential to ignore even for internal use, and they will inevitably be deployed by companies in the ecosystem in which oil and gas businesses operate. Delaying action to a later date is simply untenable as a strategic response.
Instead, companies should do the following:
REVISIT YOUR CLOUD STRATEGY.
These tools are compute intense and may be quite costly at scale. They only really work via cloud computing. If your organization does not yet have a cloud strategy that enables your workforce to even test these tools, you should start by getting your cloud game on.
CAUTIOUS FIRST DEPLOYMENTS.
Deploy these tools in areas that are not mission critical, such internal chat forums in HR, for example, or as an on-demand very simple coding utility for use by end users.
BOLSTER CYBER DEFENSES.
These tools will create pressure on cyber defenses. Conduct a review of your cyber capabilities and capacities to deal with an increased volume of activity.
CONSIDER BOT DETECTION.
Recognize that whoever you’re interacting with, both internally and externally, is increasingly likely to be a machine. Deploy bot detection software to filter all correspondence so that you do not allocate scarce human talent to respond to bot-generated content (unless, of course, that content is from a legitimate source, such as a regulator or investor).
This article has taken me almost three hours to research and write. But if you pass it through a bot detector, it comes out 100% human and less than 10% bullshit.
As for the plagiarist who ripped off my intellectual property and refused to engage, I merely contacted the Chief Legal Officer at the same company to point out the issue, and he promised to ‘take care of it’. A day later the article was gone, I had an apology in hand, and an implausible explanation (‘I have no idea how that could possibly have happened’). I didn’t press the matter any further.
Check out my latest book, ‘Carbon, Capital, and the Cloud: A Playbook for Digital Oil and Gas’, available on Amazon and other on-line bookshops.
You might also like my first book, Bits, Bytes, and Barrels: The Digital Transformation of Oil and Gas’, also available on Amazon.
Take Digital Oil and Gas, the one-day on-line digital oil and gas awareness course on Udemy.
Take the one-hour Digital for the Front Line Worker in Oil and Gas, on Udemy.
Biz card: Geoffrey Cann on OVOU
email: [email protected]