If Anyone Builds It, Everyone Dies, by Eliezer Yudkowsky and Nate Soares debuted a few weeks ago to much initial fanfare. However,
I stand by my prediction that this book will have no effect on policy or the direction of AI research. It will soon be forgotten instead of widely debated.
Intellectuals and books impart much less influence compared to decades ago. Everything is too saturated and attention spans are too short or fractured. Today, books have to compete with Netflix, TikTok/YouTube ‘shorts/reels’, Substack/blogs, and social media for attention. For these reasons, books and documentaries alike are a waste of time for changing minds given the amount of time that goes into it.
Policy makers care about affecting sentiment in the short term, not long-term or abstract problems like AI risk. Additionally, the topic is unlikely to resonate with the public, and the complexity will go over the heads of policy makers despite the best attempts of the authors to explicate the issue in layman terms, especially given how dumbed-down discourse has become in recent years.
Maybe 30 years ago, such a book would have merited serious consideration by Congress or other elites, but not anymore when the only objective is point-scoring against the opposing team. It’s also harder to spin a sentimental or human interest angle on AI risk compared to issues such as crime, poverty, or education. Unlike, say, terrorism, AI risk is too abstract, and it’s hard to attach vivid imagery to AI risk compared to terrorism deaths.
Even the eye-catching promise of “everyone dying” will be met with indifference, as this is how all issues are framed (e.g. if global warming isn’t stopped, millions will die, or warnings in 2015-2017 of how ‘Trump would destroy democracy’). The public and policy makers have become cynical or deaf to such empty threats or marketing when said consequence never happen as promised (e.g. decades later, still no coastline flooding, and democracy is more or less still intact).