US think tank: AI could cause nuclear war by 2040

Artificial intelligence (AI) could potentially cause nuclear war by 2040, according to a research report by a US think tank.

A paper by the non-profit think tank Rand Corporation warns that AI could undermine geopolitical stability and eliminate nuclear weapons as a deterrent.

Rand researchers warn that while nuclear strikes have maintained peace for decades as a mutually assured destructive strategy, guarantees of stability could be eroded if AI and machine learning (ML) determine military action.

In a paper based on a series of workshops with experts, the researchers said AI could enable human actors to make lethal decisions in the future.

For example, improvements in sensing technology could destroy retaliatory powers such as submarines and mobile missiles. AI could also entice other countries to gain bargaining power by launching a pre-emptive strike against other countries, even if they are unwilling to attempt an attack, the researchers said.

In a related paper, Andrew Lohn writes: “Some experts are concerned that an increased reliance on artificial intelligence could lead to new types of fatal mistakes. There may be pressure to use AI before it is technologically mature, or it may be vulnerable to an adversary's overthrow. Therefore, maintaining strategic stability for decades to come can be extremely difficult. All parties should participate in the agency's efforts to limit nuclear risk."

The report highlighted the dangers of using AI to make military decisions, not the threat of autonomous drones and so-called 'killer robots'. The researchers pointed out that the 1983 nuclear war warning incident was a warning about the post-Cold War development of AI.

In 1983, former Soviet Army officer Stanislav Petrov discovered a warning on a computer that the United States had fired several missiles. The warning turned out to be a malfunction. Petrov, who died late last year, was credited with saving the world from nuclear destruction.

“The link between nuclear warfare and artificial intelligence is not new,” said Edward Geist, a policy researcher, co-author of the report. Indeed, the two histories are intertwined. "Most of the early development of AI was done in support of military efforts or with military objectives in mind."