AI leaders reportedly weigh catastrophe scenarios as Nadella warns of risks

Leaders at major artificial intelligence companies are reportedly comparing “day after” scenarios, while Microsoft chief Satya Nadella says society should assume every AI model is compromised from a cybersecurity perspective, ERR’s R2 Portal reports in a commentary by Kristjan Port.

Port recalls a 1960 paper by mathematician Heinz von Foerster and colleagues, published in the journal Science. Based on an analysis of humanity’s development, the authors predicted a doomsday on November 13, 2026, when population growth would reach a point at which people could no longer fit on Earth.

The prediction was not religious and did not involve a flood, aliens or the planet breaking apart. It was the result of projecting population growth forward. The authors used technical terms such as “singularity” and “infinity” to describe what that projection implied.

The prediction did not come true because humanity changed its growth trajectory. Education, urbanisation, contraception and changing views of family helped shape population growth. People adapted their behaviour as they recognised the world’s limits, a shift reflected today in record-low birth rates.

Port says the paper’s authors intended to provoke discussion: if a social system does not change, mathematical extrapolation can lead to absurd conclusions. A model may describe the past accurately yet fail to predict the future when the system changes.

Intelligence is central to adapting to change, Port writes. There was once optimism that artificial intelligence would help solve increasingly complex problems, but that confidence has given way to anxiety. In a difficult period, the old doomsday prediction has found a new audience, while the fast and uncontrolled development of AI reflects doubts that existing judgement and agreements will be enough.