The Waymo effect

The focus of the September issue of Nuclear News issue is on new technology, and the stories within are reflective of the broader wave of innovation sweeping the industry-a wave driven in no small part by the head-jerking progress of AI and its pervasion into our economy and way of life.

Something caught my eye a few days ago: an ad by Waymo touting its careful intentionality in deploying its fleet of driverless vehicles in the U.S. "Now, the Waymo driver is safer than a human one," says the soothing female voiceover, "statistically 10 times safer in the cities we serve."

It made me realize that most if not all safety-related applications of AI will at some point- usually after a lot of trial and error-reach a state of "too good to not use." At that point, people become the weakest link in the safety chain. While the first Waymo ride can be a cognitively dissonant experience, I think we will be surprised at how easily we will all get used to seeing the road ahead behind an empty driver's seat, assuming the safety case holds up.

I believe we are on the same journey with advanced nuclear reactors. In fact, I would argue that the widespread adoption of new nuclear energy technology will be nearly impossible without the broad use of process automation, machine learning, digital twins, and ultimately autonomous operation to maintain safety while achieving high levels of operational efficiency.

This means we must prepare our fellow humans to confront their nuclear "Back-of-the-Waymo" moments when they come. There is no shortage of concern about AI out there, from nuclear engineering students who worry about its impact on the future of the nuclear profession (and their professors who fret over their students' hand calculation fluency) to the larger dystopian outcomes that seem to pervade the podcast-o-sphere. But if you work in the field, it's hard not to be optimistic about the benefits AI can bring to the nuclear sector, and ultimately humanity-benefits that don't need to come at the expense of human agency.

Yes, we can travel in Waymos without undue fear, but we must bring our fellow humans along for the ride. At ANS, we are playing a role in ways we never expected. When I joined the Society, our digital publications presence was almost nonexistent-roughly 450 copies of NN were downloaded each month by members. We created Nuclear Newswire to make our journalistic content more accessible to a wider swath of readers, and over the subsequent five years, NN article views grew from zero to over 100,000 per month at a steady 20 percent annual rate. But in the last year, something weird happened: Our page views effectively tripled to over 300,000 a month. Why such a drastic increase? Because more people are researching nuclear topics using large language models like ChatGPT and Claude, which in turn are coming to us for answers. We can see it in our declining "time per session" stats. In effect, we have become AI's nuclear technology instructor.

That's a strange kind of responsibility: teaching a technology that will in turn teach millions of people who will never read NN directly. But if AI is going to be our profession's newest student, I'd rather ANS be the teacher than leave it to chance. Turns out our own Back-of-the-Waymo moment arrived sooner than I anticipated-but I think we're ready for it.