Changing the world for the worse

A former effective altruist working in AI safety has published a viral essay warning that the movement's efforts may be accelerating the very risks it seeks to mitigate. The author, who worked at an EA organization influencing government projects worth tens of millions, describes how the FTX collapse exposed the dangers of unchecked ambition at global scale. More critically, they observe that the AI safety community's work—inspired by Nick Bostrom's 'Superintelligence'—has inadvertently fueled the capabilities race between OpenAI and Anthropic.

The essay provides insider perspective on how young EA professionals gained influence over millions of dollars before age 25, creating a 'heady feeling' of unlimited potential. The author notes that people close to them were 'thrown into turmoil and depression' when they realized their AI safety work had primarily accelerated AI capabilities development rather than ensuring safety. This realization came alongside the OpenAI-Anthropic competition that Bostrom's framework would consider 'just about the worst thing that could have happened.'

The author ultimately quit their EA position, not through dramatic conviction but through burnout and disillusionment. They now argue that the most responsible approach for those taking AI existential risk seriously might be to 'stop working on AI at all.' The core thesis challenges the EA assumption that large-scale intervention is inherently positive, suggesting instead that uncertainty about outcomes makes such ambition dangerously hubristic—especially when dealing with transformative technologies like advanced AI systems from leading labs.