Global artificial intelligence (AI) experts have assessed that there is a greater than 20% chance AI could cause significant human and economic harm within the next five years.
According to a report by Axios on July 27, researchers from the Massachusetts Institute of Technology (MIT) Future of Technology and the University of Queensland in Australia surveyed 272 AI experts from 37 countries to evaluate AI risks.
The researchers assumed that companies and governments would continue to develop and utilize AI in the same manner as they do now. They asked experts about the likelihood of 24 different risks leading to catastrophic events between 2025 and 2030.
Catastrophic events were defined as those resulting in over one million deaths or economic losses exceeding $100 billion. This also included severe societal impacts, such as the collapse of democratic order.
The survey found that the likelihood of 'dangerous AI capabilities' causing a catastrophic event was the highest at 21.5%. The potential for large-scale damage from cyber attacks using AI and the development and use of AI in weaponry was also assessed at 21.0%.
The risk of power and benefits from AI being concentrated among a few was rated at 18.0%. The competition between nations and companies in AI development, which could weaken safety measures, was assessed at 16.6%, while the risk of misinformation or misunderstandings spreading was rated at 12.8%.
Among the 24 risks evaluated, 18 were deemed to have a greater than 10% chance of leading to catastrophic events.
However, the researchers noted that these risks could overlap. They cautioned against simply adding the individual risk percentages or interpreting them as the overall probability of AI causing a disaster.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.