Learning to Live with (de)Generative AI

A quick trial of ChatGPT reveals its considerable strengths and its worrisome weaknesses in generating energy talking points.

One of my many contacts asked me a rather simple question earlier this week: Why does Greenpeace not like LNG? My contact is from Germany, and presumably has lots of family members and friends there who have been living through Germany’s recent natural gas energy angst.

As a thought experiment I wondered how a modern digital solution like ChatGPT might answer the same question, so I asked it the same question.

ChatGPT did not disappoint, but neither did it impress. I would be cautious relying on ChatGPT to help with understanding the complex world of energy security. This little episode is a fine illustration of the perils of this kind of technology at this early stage of its life:

• ChatGPT’s grasp of the technical facts around LNG, a 60 year old technology, is weak. That does not bode well for its understanding of more modern and complicated technologies.

• ChatGPT’s responses are quickly generated, which is value, but we need to be very sure we are asking the right questions, and we take the time to quiz the responses to see if interesting and relevant contradictory information is being unintentionally withheld.

• Many in society lack the patience, the skepticism, and the inquiry skills to push back on the narratives produced by generative AI. Fortunately, that’s not the oil and gas industry. We’re a skeptical bunch.