Global artificial intelligence (AI) experts assess that there is an average probability of over 20% that AI could cause "catastrophic outcomes," such as killing millions of people or causing hundreds of trillions of won in economic losses within the next five years, a survey showed.

According to U.S. online media outlet Axios on the 27th (local time), these findings came from a Delphi method survey conducted late last year among 272 AI experts worldwide by FutureTech, an interdisciplinary research group at the Massachusetts Institute of Technology (MIT), and researchers at the University of Queensland in Australia.

The report on the results was released in June of this year, and MIT distributed a press release regarding it on July 20 (local time).

The researchers presented experts with 24 types of AI risks and asked them to evaluate the probability and severity, sector and actor vulnerability, actor responsibility, and overall concern for each risk type, under the assumption that companies maintain their current trajectories.

"Catastrophic outcomes" were defined as cases "causing more than 1 million deaths," "causing economic losses exceeding 100 billion dollars (145.5 trillion won)," or "causing civilization-scale intangible impacts, such as the collapse of democratic norms."

Synthesizing the experts' opinions, 18 out of the 24 risk categories were evaluated as having a probability of 10% or higher of occurring between 2025 and 2030.

Among them, particularly high-probability risks included "AI possessing dangerous capabilities" (21.5%), "cyberattacks, weapon development or use, and massive casualties" (21.0%), "concentration of power and unfair distribution of benefits" (18.0%), "competitive dynamics" (16.6%), and "disinformation or misleading information" (12.8%).

Axios explained, "The behavioral patterns of AI and public opinion surrounding it can shake markets, instantly alter public discourse, and even lead to catastrophic scenarios."

On the other hand, it added, "AI has explosively boosted productivity, and those working in the AI field highly value the potential of this new technology for enhancing productivity and automating tasks."

It also noted, "Researchers emphasize that while catastrophic risks from AI remain uncertain, they are important enough for relevant parties to recognize as challenges that must be overcome."

(Photo: Yonhap News)