The sentence is our verdict on one argument in Nothing Went Foom!, an AI-made answer song to I’m Upping My P(doom). The song lists predicted disasters that never came: the mass starvation forecast in The Population Bomb in 1968, the millennium bug, OpenAI’s decision in 2019 to hold back GPT-2, the 2023 letter calling for a six-month pause. Its conclusion: the fears were overblown, so accelerate.
Against a risk that is expected to arrive late and fast, “nothing has happened yet” is weak evidence. FOOM, the scenario of explosive self-improvement, is a claim about what happens once systems can improve themselves; that it has not happened is no argument that it cannot. The International AI Safety Report 2026 supports caution without supporting panic: today’s systems lack the capabilities for a loss of control, it finds, but they are improving in the relevant areas.
The song’s own examples make a better case than the one it sings. The global famine did not come – most accounts credit the plant breeders and farmers of the Green Revolution – and the millennium bug was fixed by engineers rewriting code. These were risks met by work, not by moratoriums. And there is the cost of waiting: if AI can speed up medicine, every year of delay has victims who never appear in a risk model. Those who want to stop must count what stopping costs, just as those who build must count what building risks.
Read more: Nothing Went Foom! · A Field Guide to Nothing Went Foom! · Related: FOOM · Accelerationism · AI Pause · p(doom)
Sources: International AI Safety Report 2026
Leave a Reply