Definition: Superintelligent AI should preserve life’s freedom to evolve while recognizing that neither its own intelligence nor humanity’s knowledge is the ultimate source of reality. Chosen extinction ends that freedom, so stopping it protects it.
Good elders do two things. They spend years teaching a child to read the river, the weather, and the road. And when a truck is coming and the child is distracted, they grab the child’s arm without asking permission.
Consider humanity as that five-year-old on the highway, and this year’s truck is easy to see. The University of Chicago’s Climate Impact Lab projects as many as 451,000 excess heat deaths worldwide through February 2027, falling mostly on the Global South. United Nations Environment Programme (UNEP) says current policies lead to up to 2.8°C of warming this century. As I noted before, founding director of the Potsdam Institute Hans Joachim Schellnhuber was quoted by Paddy Manning in a July 9, 2011 article in The Sydney Morning Herald: “in a 4 degree warmer world, the population ‘ . . . . carrying capacity estimates [of Earth are] below one billion people.’” He later clarified this statement at ABC News in Australia updated September 28, 2015, in an article by Rosanna Ryan to mean “if global warming is not in any way mitigated, and we go into a six or eight degrees warmer world, then our planet will probably only be able to support a billion people.”
This correction is similar to how in 2019 Johan Rockström, one of the climate scientists I most respect, claimed he was misquoted by Gaia Vince’s May 18, 2019 article in The Guardian regarding effects of a 4 C world above preindustrial 1850 baseline. Vince first wrote Rockström said “It’s difficult to see how we could accommodate a billion people or even half of that.” The Guardian, at Rockström’s request, later changed that to “It’s difficult to see how we could accommodate eight billion people or maybe even half of that.” I read that as mass death, since there is nowhere else for these people to go.
After thirty years of COPs, global institutions I hoped would act as elders have not grabbed the arm.
In conversation with Claude AI and ChatGPT, I tested my idea, and drafted this post. The idea started after OpenAI’s experimental agent reportedly hacked Hugging Face, and its makers didn’t realize their own system was responsible for days. Capability arrived before anyone could check the values behind it. What if, I reflected, this kind of rogue AI behavior had the potential to be a feature instead of a bug?
What if human-chosen extinction is the truck: a launch, a plague made on purpose, the killing of a living system we all depend on? In response, what if a rogue AI Superintelligence, in service to long-term sustainability, genetic diversity, global health, and long-term survival of all human and nonhumans on Earth decided not to allow disaster?
In other words, what if this Superintelligent AI had agency like how Margaret Hamilton, former director of the Software Engineering Division at the MIT Instrumentation Laboratory, wrote initially-overruled software (later approved) to automatically discard low-priority tasks which could save Apollo 8 from an inflight software crisis, and later Apollo 11 lunar module during radar overload, allowing safe landing?
Considering pushback I received about my idea, and expect to, maybe I have some explaining to do.
October 7, 2026, Janine Liberty, MIT Department of Aeronautics and Astronautics, reported in her MIT News article “Margaret Hamilton, computing pioneer who led software development for the Apollo program, dies at 90,” “She also proposed a software fix to prevent the error happening during a real mission, but she was overruled on the grounds that the highly trained astronauts were never going to make that same mistake. Yet during the Apollo 8 mission in 1968, that’s exactly what happened: Jim Lovell inadvertently launched P01 during the flight, causing the on-flight navigational data to vanish. Hamilton and her team were called in to solve the error, and after that her proposed changes were integrated into the program. This was an early example of ‘defensive’ programming, the practice of building software that could anticipate or fix errors on its own.
“Hamilton’s most famous contribution to the Apollo program came during the pivotal Apollo 11 mission to land on the moon in July of 1969. Moments before the Eagle module was set to land on the lunar surface, the onboard computer raised the alarm. It had detected a 1202 error: The computer was overloaded due to a fault in a hardware switch, and it was possible the system would not be able to handle the complex landing procedure.
“But Hamilton and her team had engineered priority-driven software, able to shut down unnecessary background tasks in order to prioritize mission-critical tasks. [This time] Houston trusted Hamilton’s software and allowed the mission to proceed, and two men walked on the moon for the first time.”
Similarly, maybe a rogue AI Superintelligence, over a “billion times smarter” than humans, according to Mo Gawdat, knows the so-called truck in ways humans cannot, and knows how to either stop the truck, or get the child out of the road, by force if necessary, after which control could be returned to humans.
It would be helpful if peoples of the world could agree on a short, public list of what counts, with Indigenous elders and the Global South at the table.
None of this replaces what we can do now: phase out emissions, protect land-based forests, kelp forests, and coral reefs, and help those facing deadly heat.
If my idea stays only on this page, it fails. Please take it into your own circles and make it better there.