Preservation after death for future revival could ease some AI accelerationists’ sense of urgency and make a pause more likely, says Nectome co-founder Aurelia Song
Preservation after death for future revival could ease some AI accelerationists’ sense of urgency and make a pause more likely, says Nectome co-founder Aurelia Song
On November 9, Aurelia Song published an essay titled “Preservation Can Save the World!” on the LessWrong forum. Starting in early 2027, Nectome plans to preserve the bodies of terminally ill people immediately after death.
Supporters of faster AI development, or “accelerationists,” argue that pausing development would cost lives because AI will learn to cure diseases. “We believe any deceleration of AI will cost lives,” investor Marc Andreessen wrote in “The Techno-Optimist Manifesto”. Song believes that, for many accelerationists, this argument stems from love for those close to them: “They do what they do out of love.”
Song once reasoned the same way. She went to the Massachusetts Institute of Technology to work on mind uploading, but in 2007 she judged the methods too primitive and turned to AI. She estimated that mind uploading would take 70 years to develop without AI and 20 with it. At 60 million deaths a year, that 50-year difference would mean about three billion lives saved. AI risk did not worry her: she assumed AI would act in humanity’s interests by default, and the fate of future people felt more abstract than that of her living loved ones. In 2015, she learned that AI could also act against humanity’s interests, left that work, and turned to preservation, which she believes saves lives without that risk. Since preservation helped her change her mind, Song writes, it could help others do the same. Her argument is:
“Preservation offers a much safer option to accelerationists who are rushing toward AI in the hope of saving loved ones who have only a few years left to live. Those people can be preserved, so there is no longer any need to rush.”
The argument rests on two assumptions: that enough of the brain’s structure can be preserved today to allow revival, and that the time when AI will learn to cure diseases is unknown.
Song bases her second argument on anthropologist Ernest Becker’s book “The Denial of Death”: people defend themselves against the thought of death through denial (“death does not exist”) or acceptance (“death is good”). Denial distorts their view of the world; acceptance distorts their morality. Preservation, Song argues, lets people “do something about it.” When people plan for 300 years rather than 60, many one-off decisions become recurring ones, and cooperation often becomes more advantageous in repeated interactions. She believes humanity would be better able to make AI safe with that longer time horizon.
In her six-step plan, Song expects Nectome to preserve a well-known person in 2027 and the idea of preservation to reach world leaders and accelerationists. She believes this would make a pause more likely: a substantial share of accelerationists would agree to one once they were less worried about the imminent deaths of people they know. She writes that the race has other causes besides love for people who are alive today.
In August, neuroscientist Ariel Zeleznikow-Johnston proposed brain preservation as a safe alternative to betting on faster AI development: the less inevitable death becomes, the less reason there is to take risks. But in the same essay, he wrote: “I don’t think the prospect of abolishing death soon is the main thing driving AI companies to take risks.” The leaders of major laboratories have decades of life ahead of them, and money and power seem more plausible motives. In a comment on Song’s essay, he added that the justifications for acceleration seemed to him to be after-the-fact rationalizations, and that accelerationists would still push recklessly ahead even with preservation available. He qualified this as “just a feeling.”
A commenter suggested surveying accelerationists to ask whether inexpensive, reliable preservation would make them feel less urgency. Song asked how to do that.