In an upcoming book co-authored with Nate Soares, Eliezer Yudkowsky doesn’t pull any punches, promising in the title that if anyone builds superhuman AI, that ‘everyone in will die.’ But It’s more likely that rather than everyone dying, this book will go in the dustbin of history. Like most non-fiction books, the putative objective of his book is a call to action and to raise awareness–in his case, about AI risk. However, two things: the public and policy elites are already abundantly aware of AI risk. Second, the likelihood of any book affecting discourse or policy is tiny.
About the first point, AI risk is a recurring theme of the chattering classes. The New York Times, Time Magazine, Newsweek, and The Atlantic have published many articles over the past couple years entertaining the possibility of AI risk. In May 2023, the NYTs published “AI Poses ‘Risk of Extinction,’ Industry Leaders Warn”. In May 2024, The Atlantic published “The Big AI Risk Not Enough People Are Seeing”. In early 2023 Yudkowsky went on the ‘podcast circuit’ to warn the public about such risks; these appearances generated millions of collective views.
In March 2023, he published in Time Magazine “The Only Way to Deal With the Threat From AI? Shut It Down,” which introduced the concept of AI risk to a large, non-tech-savvy audience and generated considerable discussion. In September he was also profiled by Time. Recapitulating this risk will likely not move the needle much, as almost everyone is abundantly aware of it.
About the second point, the probability of a non-fiction book having a tangible effect on public discourse, let alone policy, is tiny, even if the book has famous authors and gets considerable media attention. There are simply too many books and not enough time, readers, or attention for each book to get the consideration merited of it. And even among the books that are read, the public’s attention is already divided among competing issues, like climate change or China. This is made worse by social media, Netflix, smartphones and other distractions or diversions which didn’t exist 20+ years ago.
Even if a book is initially a hit on social media, it tends to fade fast. For example, Richard Hanania’s The Origins of Woke: Civil Rights Law, Corporate America, and the Triumph of Identity Politics, published in September 2023, initially gained a lot traction on social media and even the mainstream media, such as being reviewed by The Atlantic, but hardly anyone talks about it anymore. Compare to The Population Bomb, a 1968 book co-authored by former Stanford University professor Paul R. Ehrlich, which actually was debated by the public and academia and implemented by some policy elites, such as India (albeit with bad results).
Additionally, the era of proactive national government is over. The 20th century era of impressionable policy elites who were receptive to the idea of ‘big government’ affecting change, came to a close by the new millennium. Such massive undertakings (e.g. the Works Progress Administration, the Manhattan Project, ‘the space race’, the ironically-titled ‘Great Society’ programs, and others) signified optimism about the power of government, even if these programs were bloated or otherwise ineffective. But this has given way to cynicism or caution.
Policy elites, on either side of the aisle, now care more about affecting sentiment and getting applause, than creating long-standing change at the government level. For example, during Covid, rather than a sweeping national response to contain the virus, as seen in other countries, it was left it up to the states to decide how few or many restrictions to enact. Compare to post-9/11, in which entire agencies and laws were created in response to the threat of terrorism. In light of these facts, it’s hard to imagine the federal government marshalling a national response to AI risk either.