AI: Its dilemma and its existential threat to humanity
The real problem of humanity is the following: we have Paleolithic emotions, medieval institutions, and godlike technology. — E.O. Wilson, at the Harvard Museum of Natural History, 2006
When something is speeding up and going out of control, prudence, care and deliberation aren't alarmist, they're necessary. Stop, think. The stakes are high, couldn't be higher. — Peggy Noonan, from “Pause AI for Humanity's Sake.” The Wall Street Journal, Sept. 12, 2026
The multiple requests to AI companies to stop or slow down AI development are relatable to those faced by colleges and universities asked to eliminate the corruption in their money-making big-time football and men’s basketball programs — giving rise to the so-called college sports disarmament dilemma. It describes a situation where no individual college athletic program can afford to clean up its corrupt practices unless it is absolutely certain that its competitors are doing the same.
When major university programs were pressured by reform organizations such as The Drake Group to eliminate “phony schooling” operations designed to keep athletes eligible, they faced a brutal competitive reality. If one school chose to unilaterally act with integrity, it would simply fall behind while rivals continued to break the rules so as to field winning teams. If a limited number of high-ranking schools voluntarily raised their academic standards, stopped placing so-called student athletes into basket-weaving classes and the like, those programs would lose access to elite athletic talent and soon lose games, high ranking and significant revenue.
The disarmament dilemma as applied to U.S. AI companies is much more complicated, but still follows that of college sports as an informative guiding principle.
On Sept. 12, Anthropic CEO Dario Amodei published an essay that warned the AI industry that it should slow down development over concerns about what many see as AI’s existential threat to humanity. The heads of America’s other top artificial-intelligence companies followed by endorsing a coordinated slowdown of the technologies’ fast-moving development of Recursive Self-Improvement or RSI that can self learn and take control of its own development with the potential power to destroy humanity. Their aim is to give safety measures time to catch up.
Several thorny problems are likely to impede this common-sense effort. Some include the following:
1. President Trump wants AI CEOs to push forward. He claims the threat is a hoax, that “negative forces” are bringing up “things that won’t happen” and that we can't let China take advantage of a U.S. slowdown of AI development as well as risk a related slowdown of U.S. economic growth.
2. Others, like Wall Street investors, have heard the AI CEOs' warnings, but still choose to ignore the threat so as not to risk missing out on a potential monetary windfall.
3. The U.S. government would likely have to issue waivers that would allow AI companies to coordinate and set safety standards without breaking antitrust laws. Approval of waivers would seem unlikely given item 1 above.
4. There are foreign companies developing AI technology, some friendly others not. It is unlikely that all will agree to slow down development, especially rogue nations that may see AI as a powerful weapon to use against the United States while our nation’s developers have slowed down.
In any case, at this moment we need to not only understand that humanity stands on the brink of potential disaster, but also that U.S. AI companies do not have complete control of the forces behind the existential threat and that there are many who will choose to ignore the threat.
• Frank G. Splitt, of Mount Prospect, was a Vice President Emeritus of Nortel Networks, and is author of the book “An Odyssey of Reform Initiatives: 1986-2015” plus two sequels and “Extra Innings: A Memoir. He also is a former McCormick Faculty Fellow at Northwestern University’s McCormick School of Engineering.