Will Artificial Intelligence ‘Kill Us All?’

Steve Hayes is joined by Jonah Goldberg, Mike Warren, and Klon Kitchen to discuss the resignation of an Anthropic researcher over concerns that AI may “kill us all,” yet another flip-flop by the Trump administration on Ukraine, and President Trump’s $5,000 bribe to voters during the midterm convention. The Agenda: —Pacing the AI frontier —AI and foreign policy Show notes: Steve Hayes is CEO and editor of The Dispatch, based in Annapolis, Maryland. Prior to co-founding the company in 2019, he worked at The Weekly Standard for 18 years, covering Washington, politics, and national security. Steve is the author of two New York Times bestsellers. He also worked as a contributor at CNN and Fox News, and currently serves as a political analyst at NBC News. When Steve is not focused on The Dispatch, he’s probably traveling with his family, grilling, or riding his mountain bike. Jonah Goldberg is editor-in-chief and co-founder of The Dispatch, based in Washington, D.C. Prior to that, enormous lizards roamed the Earth. More immediately prior to that, Jonah spent two decades at National Review, where he was a senior editor, among other things. He is also a bestselling author, longtime columnist for the Los Angeles Times, commentator for CNN, and a senior fellow at the American Enterprise Institute. When he is not writing the G-File or hosting The Remnant podcast, he finds real joy in family time, attending to his dogs and cat, and blaming Steve Hayes for various things. Michael Warren is politics editor at The Dispatch and is based in Washington, D.C. A former on-air reporter at CNN and a senior writer at The Weekly Standard, he has spent more than 15 years covering lawmakers on Capitol Hill, the White House and the executive branch, the conservative movement, and political campaigns. When Mike is not reporting, writing, editing, and podcasting, he is probably spending time with his wife and three sons. The podcast transcription below was generated using artificial intelligence and may include occasional errors or misspellings. Steve Hayes 0:00 [instrumental music] Welcome to the Dispatch Podcast. I'm your host, Steve Hayes. On today's podcast, we discuss the end of humanity. No, really. Last week, Anthropic researcher Jacob Coxen quit the leading AI firm, arguing that, quote, "The people building AI earnestly believe that it could kill us all by the end of the decade," unquote. His comments, supported by other leading AI researchers, kicked off a feverish debate about the future of AI and humans. We take a careful and sober look at that debate and the issues involved. Then the Trump administration over the weekend blamed Ukraine for prolonging the war Russia started nearly five years ago, and for creating a diesel fuel crisis with its attempts to defend itself. What to make of those claims? And finally, not worth your time, the president stakes the midterms on a five thousand dollars bribe? I'm joined by Jonah Goldberg, Dispatch co-founder, Mike Warren, politics editor at The Dispatch, and Klon Kitchen, who for fifteen years served as an intelligence officer focusing on counterterrorism, cyber operations, and covert actions. Klon Steve Hayes 1:12 was National Security Advisor to former Senator Ben Sasse and played a pivotal role in the US Cyberspace Solarium Commission, crafting the strategic blueprint for America's cyber defense. Most significantly, Clawn is the former author of The Dispatches Tech and National Security newsletter, The Current. Let's dive right in. [instrumental music] Welcome, gentlemen. The latest debate on AI started with a tweet from Jacob Coxen, an AI researcher from Anthropic who had spent three years at OpenAI before that. Coxen tweeted last week that he was quitting the company, Anthropic, because of his concerns that the race to win on, quote, "Self-improving superintelligence has the companies acting irresponsibly." As of Monday morning, his lead tweet in the thread has one hundred and seventy-one million views. A couple of brief excerpts from his thread here. "I resigned from Anthropic today. I spent the last three years doing pre-training research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. Do not underestimate the power of this technology. These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources. We've all witnessed the progress in each of these domains, and progress is not slowing. The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible, but I hear the same people express fear privately. No other human activity poses this level of danger." Klon, immediately after he tweeted that, several other senior people in AI safety research echoed his claims, including Evan Hubinger, a team lead at Anthropic, Alex Turner, a former research scientist at Google DeepMind, and Jakub Pača-Pachunks-Chucksky, chief scientist at OpenAI, urging extreme caution. Clawn, you spend a lot of your time these days looking at AI, thinking about AI, talking about AI with the people driving these developments. Steve Hayes 3:38 When you hear this, "The people building AI earnestly believe that it could kill all of us by the end of the decade," Steve Hayes 3:47 what do you think? Klon Kitchen 3:48 Well, I think Jacob is accurately representing a significant portion of the AI safety group out there. And look, there's enough capability that's been on display, which I can talk about in detail, and enough concerning developments to continue to justify some of these concerns. But not everybody feels equally alarmed by what Jacob is highlighting. Let me just begin with kind of some of the presuppositions behind what he laid out. The particular advancement that is driving what he wrote about is what's called recursive self-improvement, RSI. It's the idea of AIs building themselves, right? And so the, the fundamental concern is, up until this point, as we have been building artificial intelligence, Interactive transcripts are available to Premium members. Join to follow along with the audio. Join Now