Posted on September 20, 2023 by Stephen Gidiere
As far as I know, this headline is false. But is it so hard to imagine? EPA actually maintains an “artificial intelligence inventory” that claims EPA has developed “a proof-of-concept to improve enforcement of environmental regulations through facility inspections by the EPA and state partners. The resulting predictive analytics showed a 47% improvement of identifying violations of the Resource Conservation and Recovery Act.”
And EPA undertakes many regulatory tasks that could utilize artificial intelligence to synthesize large data sets. EPA receives large amounts of emissions and discharge data under the statutes it administers. EPA performs risk assessments with large data sets. EPA and its scientific advisory boards shift through and analyze scientific data on risks to human health and the environment in the academic literature.
It seems inevitable that the agency will be faced with deciding whether and how to incorporate artificial intelligence tools into these tasks. And is it that far-fetched that ChatGPT could be asked to write all or a portion of a preamble (or may be already has been)? “Write a summary of the available technologies for controlling the emission of sulfur dioxides.” Or: “Write a history of EPA’s regulation of ozone.” Or: “What are the global impacts from the release of greenhouse gases into the atmosphere.” Of course, this would require an accurate and robust set of data, and, if you have spent any time with EPA’s ECHO database or submitted an information correction request to EPA under the Data Quality Act, you would know the shortcomings in that department.
So let’s hope my false headline remains false. But if EPA does make such a proposal, let’s request an AI response: “Write a set of comments explaining why EPA cannot use artificial intelligence to identify enforcement targets.”