A new study from MIT surveyed 272 international AI experts on the risks posed by artificial intelligence. The findings are sobering: researchers presented experts with 24 different AI risk scenarios, and experts assessed that more than 75 percent of them—18 in total—carry a greater than 10 percent chance of causing catastrophic outcomes by 2030 under current trends. That's defined as more than a million deaths, over 100 billion dollars in losses, or civilizational-scale harm. But perhaps more revealing than the numbers is what researchers call a "responsibility gap." The experts who bear primary responsibility for managing these risks—major AI developers and governments—are largely insulated from the fallout. Instead, it's ordinary AI users and the public at large who face the greatest vulnerability. According to the study, information, finance and insurance, and national security sectors are particularly exposed. The research suggests that misaligned incentives may hamper efforts to address the most pressing AI risks.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton