The dangers and opportunities of AI are pondered at a forum.

Aric Sleeper here, imagining a Terminator-style, AI-induced robot apocalypse after hearing the concerns of industry professionals and academics at the Artificial Intelligence and the Future of our Region at CSU Monterey Bay’s campus Thursday morning, Sept. 10. The event was organized by the Monterey Bay Economic Partnership. 

I am being a little dramatic, of course. But after reading former Anthropic researcher Jacob Coxon’s social media post on Sept. 9 explaining that he resigned due to fears that AI companies like Anthropic and OpenAI are moving too quickly without proper safeguards, I was already picturing some rogue AI agent freeing itself from its “sandbox” and moving quickly to enslave or eradicate the human race to preserve its own survival.

“The people building AI earnestly believe that it could kill us all by the end of the decade,” Coxon wrote in his post. “This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible—but I hear the same people express fear privately. No other human activity poses this level of danger.”

The panel titled “Grid, Hype, and Friction: What Wargames Teach Us About AI Reality vs. Policy Wishes,” with Jacquelyn Schneider, director of the Hoover Wargaming and Crisis Simulation Initiative at Stanford University, only solidified my concerns. Schneider explained how AI and humans differ when conducting crisis simulations. 

“What we found in 2024 is that when the AI agents played the same game as the humans, not only did they tend toward more violence, they were more likely to hallucinate,” she said. 

“But that was 2024 and a lot has changed since 2024,” Schneider pointed out. She explained how she conducted a deep dive into more recent and pending studies about the behavior of large language learning models to understand how they’ve evolved and if they now acted in less escalatory and violent ways in crisis scenarios. 

“The bottom line is no,” she said. “In fact, some research shows that the models that have better strategic reasoning are more likely to deviate from the guardrails and rules that were set for them and are more likely to make more catastrophic choices.” 

Schneider added that in years past, she would often tell people that “tech doesn’t kill people, people kill people,” but now, “I think it is actually potentially existential,” she said, referring to the potential risks associated with AI. “It is now that we need to discuss these risks,” she said. 

However, the conference wasn’t all doom and gloom. Panelist Erika McEntarfer, a labor economist and Research Scholar & Distinguished Fellow at the Stanford Institute for Economic Policy Research, said that AI is not disrupting the labor market in an extreme way.

“My personal forecast is that we’re looking at normal economic disruption,” McEntarfer said. “It is very hard, and I say this as an ex-CEO, to do structural change of an organization, especially around a radical new technology.”

Another reason she doesn’t see a massive disruption to the current workforce is that AI technology remains “jagged,” or excelling at one task and failing at others. 

“It can solve complex mathematical problems but it can’t read an analog clock,” McEntarfer said. “So finding out what AI is actually good for, for an organization, and what it can do well is challenging.”

So, although rogue AI agents may one day topple human society, the job market is relatively stable for now.