In the early 20th century, factory workers — later known as the “Radium Girls” — were hired to paint watch and instrument dials with radium‑based luminous paint. They were instructed to keep their brushes sharp by shaping them with their lips. In the following years, many of these workers developed devastating illnesses, including severe bone and jaw damage, anemia and cancer, that were ultimately traced to chronic radium exposure.

For Holly Elmore, an evolutionary biologist and executive director of PauseAI US — an organization that seeks a global pause to advanced AI development — this tragedy is a stark warning for our current era.

In a talk she gave Dec. 9 for the Berkeley AI Risk Speaker Series, Elmore argues that we’re repeating this mistake with artificial intelligence by assuming we can safely play with a technology we don't fully understand. “The expectation of many people in AI safety, for many years, has been that when we got to this point, the AI, once it was aligned, would figure out the answers for us,” she says.

But Elmore warns that this approach is like clearing a minefield by walking through it. As AI capabilities grow, she says, the probability of accidents increases — and unlike minor software glitches, these could be "one-shot" events that we cannot recover from. She points to risks ranging from the automated assembly of bioweapons to the unpredictable disruption of the social and environmental systems we depend on for survival. Instead of waiting for a machine to solve its own safety problems, she contends that experiments with such high-stakes technology are too costly to continue without a pause.

“The scale of the danger really could cripple civilization or cause extinction,” she says, “and the possibility of this alone is reason enough to pursue pausing frontier AI development.”

Watch a video of Elmore’s talk.

Listen to the episode and read the transcript on UC Berkeley News (news.berkeley.edu/podcasts/berkeley-talks).

Music by HoliznaCC0.

Photo courtesy of PauseAI.

Hosted on Acast. See acast.com/privacy for more information.

An evolutionary biologist makes the case for pausing AI