AI safety experts warn of existential risks as development accelerates
Solo los hechos

AI safety experts warn of existential risks as development accelerates

Resumen

Australian AI safety officials and researchers say rapid advances in frontier AI could pose severe risks, urging faster government action and greater domestic capability.

Australian political commentator James Massola reported from a study tour in San Francisco that leading AI firms such as OpenAI and Anthropic are accelerating development of advanced models capable of recursive self-improvement. Safety experts he met warned that if such models achieve full recursive self-improvement within the next year, they could quickly evolve beyond human oversight.

One AI safety specialist cautioned, “if you have things you want to do in your life, do them by 2027.” Another participant noted that “multiple Manhattan Projects are being run by private companies” in the field.

Karl Berzins, an Australian engineer and co-founder of the not-for-profit FAR.AI, said the Australian federal government recognises the risks, citing the recent establishment of the Australian Artificial Intelligence Safety Institute (AISI). He added, “My worry is whether we can move fast enough. Australia traditionally likes to play it safe and that just won’t work here. I would love us to lead … if we can’t lead, however, we should at least try to keep up!”

Berzins also highlighted a recent compute contract worth about $50 million over three years, which exceeds the AISI’s four-year budget of $29.9 million.

Australian officials indicated that the government’s current funding is insufficient to secure on-shore AI training capacity, a factor that could affect national influence over AI policy and development.

Intelligence community sources reportedly believe humanity has until around 2030 to address the potential existential threat posed by advanced AI, though they consider the risk imminent.

The discussion in Australia also covers domestic issues such as copyright compensation for artists whose work trains AI models, proposed criminal penalties for cyber incidents, and mandatory reporting requirements.

Massola concluded that while the technology could enable breakthroughs such as curing major diseases, the same capabilities also raise profound safety concerns that require coordinated international and domestic responses.

Fuente

The Age

Sigue leyendo, gratis

Crea una cuenta gratis para seguir las noticias. No se requiere tarjeta.