Is artificial intelligence coming for your job?
Current analysis indicates you might be in danger of replacement if you belong to any of the following groups:
- Security traders
- Customer service reps
- Human beings
At this point, paranoia is understandable. Computing and automation seem to be taking over an ever-larger share of daily life.
But geologists and geo-technicians probably have less to worry about than other professionals in the energy industry. A recent study of AI’s potential job effect on various occupations showed a fairly modest impact on geoscientists.
AI broke into the news in a big way earlier this year, for two reasons. First, ChatGPT became controversial and even notorious for its use in academics. Teachers and professors complained that they often couldn’t distinguish between a student-written paper and a machine-generated paper from the AI chatbot.
Developed by research lab OpenAI and released last November, ChatGPT reportedly had reached 100 million users just two months later.
The second burst of AI news was even more startling. Several prominent computer scientists and experts warned that generative AI programs like ChatGPT could eventually become an independently thinking threat to humans.
An opinion piece in Scientific American magazine was subtitled, “Artificial intelligence algorithms will soon reach a point of rapid self-improvement that threatens our ability to control them and poses great potential risk to humanity.”
Geoffrey Hinton, a pioneer in neural networks and computing, announced that he had left his job at Google in May to warn about the dangers of AI.
The Center for AI Safety released a statement that read, “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.” It was endorsed by more than 350 computing executives, researchers and industry leaders.
Scary stuff.
Geology Still Belongs to Humans
While some paranoia at this point is understandable, it’s still paranoia.
Mrinal K. Sen is a professor of geophysics in the department of geological sciences and Jackson chair in applied seismology at the University of Texas at Austin. A specialist on seismic wave propagation, his recent research work has included physics-assisted machine learning for seismic data analysis.
“It’s not going to replace the geoscientists. It’s not possible,” Sen said.
“This will basically give us a set of tools that will help us do our jobs more efficiently and much, much faster,” he said.
For Sen, the advantage offered by machine learning and its extension, artificial intelligence, is mainly a matter of speed. He said there’s still “a huge time gap” in compiling seismic data, processing it and then interpreting it.
“I would say it’s a year and a half. Or, if you’re lucky, a year,” he noted.
He foresees a way to move from the data stage directly to the visualization stage, what he calls the ability to “jump the loop.”
“These datasets have become enormous. This is where we can see machine learning and computer analysis helping us in a big way,” he said.
Researchers in oil and gas have considered another shortcut, according to Sen, using data and visualization the way doctors now image the human body. But doctors start off with a pre-determined model. They don’t have to figure out what a body looks like.
“That sounds really appealing, but comparably the human body is well understood, the human anatomy,” Sen observed.
“It’s quite obvious the techniques that have been used in medical imaging cannot be imported (to visualizing the subsurface) in a straightforward way,” he said.
Machine learning and AI offer a new path. They can help geoscientists recognize patterns and evaluate data much more quickly, while quickly assessing millions of potential data interpretations. And now, with generative artificial intelligence, they also can help to produce new concepts.
Generative Transformers
The GPT in ChatGPT stands for “Generative Pretrained Transformer.” GPTs are artificial neural networks, loosely inspired by the biological neuron network of the human brain. They typically train on huge quantities of input using self-supervised or semi-supervised learning.
“Generative” and “transformer” are key words, since AI with generative capabilities can learn from input and generate similar but new output: new word combinations, new images, new ideas.
Edward Felten, a professor of computer science and public affairs at Princeton University, has teamed with management and business professors Manav Raj of the University of Pennsylvania and Robert Seamans of New York University to evaluate the potential impact of AI on various occupations.
Their paper describing the study, “Occupational Heterogeneity in Exposure to Generative AI,” was published in April.
In their methodology, the researchers considered 10 different AI applications, such as image generation, language modeling and abstract strategy games, and 52 human abilities, including oral comprehension, oral expression and inductive reasoning.
Then the 52 abilities were linked to more than 800 occupations using the Occupational Information Network database developed by the U.S. Department of Labor. Focusing on the applications of language modeling and image generation, the study constructed exposure scores for each occupation. Overall scoring was set to a mean of zero and a standard deviation of one.
In AI language modeling, with results linked to occupation code and job title, 41-9041 Telemarketers had the highest exposure score of 1.926, and 25-1123 English Language and Literature Teachers, Postsecondary, the second highest at 1.857.
Lowest scores went to dancers, -1.793, brickmason and stonemason helpers, -1.822, and textile and garment pressers, -1.854.
By comparison, 17-2171 Petroleum Engineers had an exposure score of 1.033, 19-2042 Geoscientists (except Hydrologists and Geographers) 0.567, 19-2043 Hydrologists 0.555, and 19-40041 Geological and Petroleum Technicians -0.027.
In AI image generation exposure, petroleum engineers had a score of 1.576, geoscientists 1.057, hydrologists 0.898, and geological and petroleum technicians 0.384. Overall in the category, 27-1025 Interior Designers had the highest score of 2.421 and 27-2031 Dancers the lowest at -3.655.
Learn to Dance?
Having an occupation that’s strongly affected by AI doesn’t necessarily mean having a job that will be disrupted by AI. Sen noted the importance of human geoscience expertise in evaluating and applying computer output.
“For geology and geophysics, we have to have domain expertise to appreciate results,” he observed.
Sen said he expects that, within five years, AI and machine learning will make a truly significant change in the way energy professionals work. He compared the change with “going from yellow pencils to workstations.”
“The challenge is that we really have to understand geology in order to understand the machine learning analysis,” he said.
In any technological transition, certain jobs and job skills will disappear. The oil and gas industry no longer hires people who use slide rules or calculate spreadsheets by hand. For the most part, nobody misses those jobs. But if paranoia is a problem and you want a job beyond the reach of AI infiltration:
Buy a new pair of dancing shoes, and brush up on your steps.