The philosophical starting point is Peter Singer's drowning child (1972): if you can save him at the cost of muddying your clothes, you must, and distance changes nothing. In the late 2000s, GiveWell compared charities by cost-effectiveness, Toby Ord's Giving What We Can asked members to pledge at least 10% of their income, and 80,000 Hours gave career advice, including "earning to give." The name was chosen by vote in 2011. Causes are ranked by importance, tractability and neglectedness.
The turn toward AI came with longtermism, a term coined around 2017 by William MacAskill and Toby Ord: positively influencing the long-term future is a key moral priority of our time, because future people could be vastly numerous. If so, reducing an existential risk by a tiny fraction is worth more than almost any other cause. In The Precipice (2020), Ord put the probability of an existential catastrophe this century at one in six, and that from unaligned AI at one in ten. The movement overlaps with the rationalists, and in small fields such as AI safety much of the funding came from its donors, such as Open Philanthropy; doomerism has one of its cradles there.
Then came the crises. In November 2022 FTX collapsed; its founder, Sam Bankman-Fried, was the movement's star donor, and in 2024 he was sentenced to 25 years in prison for fraud. A year after the collapse, in the OpenAI board crisis, two board members' ties to the movement put it in the spotlight again. The deeper criticisms vary: that multiplying tiny probabilities by astronomical values invites being mugged (see Roko's Basilisk), that it prefers philanthropy to institutional change, and, according to Gebru & Torres (2024), that it belongs to the "TESCREAL" ideological bundle. Can an expected-value calculation over trillions of future lives guide present decisions?