DOT to Use Artificial Intelligence to Draft Safety Regulations, and That Worries People
Soon, the safety regulations that keep airplanes in the sky, car passengers safe, and prevent gas pipelines from exploding might be written by Google Gemini. This worries some people in the Transportation Department, especially given the focus on quantity rather than quality, with the Department aiming for “good enough.”
The Role of Artificial Intelligence
So-called artificial intelligence has touched every aspect of our lives. Everything is AI-something these days, even though there is nothing truly intelligent about it. The term “artificial intelligence” almost always refers to large language models (LLMs), which are anything but intelligent. Instead, they excel at predicting what word (or pixel) can best be used next to another by analyzing countless similar contexts and what humans would have chosen.
Concerns Over AI Hallucinations
AI hallucinations are common, even though the phrases written by a chatbot seem to make perfect sense at first. That poor LLM has absolutely no idea about the meaning of the words it’s generating, let alone being intelligent. Despite this, LLMs are increasingly used by people who are too lazy to conduct their own research. The downside is that these individuals often have no way of knowing whether the chatbot is correct.
Worries Within the Transportation Department
This should explain why some people inside the Transportation Department are worried that the DOT now wants to use AI chatbots to draft safety regulations in their line of duty. Staffers are concerned that any failure to catch AI errors could result in flawed regulations that, at best, could lead to litigation and lawsuits. At worst, they could lead to injuries and deaths in the transportation system.
Demonstrations of AI Capabilities
During a demonstration of Gemini’s rule-drafting abilities, the LLM produced a document that lacked key text, which a staffer had to fill in. However, DOT’s top lawyer, Gregory Zerzan, doesn’t seem bothered at all. In a December meeting, Zerzan explained that he wanted Gemini because it could draft rules in under 30 minutes, rather than weeks or even months for human advisors.
“We don’t need the perfect rule on XYZ,” Zerzan told DOT staffers at the meeting. “We don’t even need a very good rule on XYZ. We want good enough.”
The Serious Implications
However, these rules touch every facet of transportation safety, and “good enough” just won’t cut it. In this case, any error could have serious consequences. It seems “wildly irresponsible”, as one staffer confessed to ProPublica. Among the rules drafted with Gemini’s help is a still-unpublished Federal Aviation Administration rule, according to a DOT staffer briefed on the matter.
Government Push for AI Adoption
Apparently, Donald Trump has requested that federal agencies adopt AI at a fast pace. However, drafting laws and regulations has not been specifically mentioned. Nevertheless, Zerzan wants DOT to be “the point of the spear” in this effort and “the first agency that is fully enabled to use AI to draft rules.”
