Debunking AI Doom: Why Existential Risk Arguments Fall Apart | Ep 12

Is artificial intelligence really on track to wipe out humanity by 2050, or are the most prominent doom arguments fundamentally flawed? In this episode of Philosophy Programs and Prompts, Casey Hart( @OntologyExplained ) and Carl Brown( @InternetOfBugs ) break down the core premises behind AI existential risk (x-risk). From Bostrom-style superintelligence scenarios and recursive self-improvement (RSI) to "hockey stick" benchmark progress and the appeal to expert anxiety, Casey and Carl analyze where the logic holds up—and where it jumps the rails. More importantly, they discuss why hyper-focusing on apocalyptic sci-fi scenarios distracts the public and regulators from the very real, immediate harms AI systems are causing right now: cyber vulnerabilities, military reliance on hallucinated outputs, non-consensual deepfakes, and corporate regulatory capture. Subscribe for more grounded, hype-free takes from two real humans who know the tech! More info about Casey:https://CaseyHart.com/ More info about Carl:https://InternetOfBugs.com/Episode Highlights:• Defining "Doom" and evaluating p(doom) numbers from a Bayesian perspective• The Core Doom Argument: Does superior intelligence guarantee global domination?• Demystifying Recursive Self-Improvement (RSI): Lines of code vs. real-world breakthroughs• Shifting focus from hypothetical extinction to tangible, current AI damages// References and Resources: // Casey’s appearance on Doom Debates https://www.youtube.com/watch?v=oxHKesSpqBM //AI Almost Started a War with China https://militarywatchmagazine.com/article/ai-us-intelligence-war-china-nuclear // Apple Research Paper on Reasoning Fragility & Contamination https://arstechnica.com/ai/2025/06/new-apple-study-challenges-whether-ai-models-truly-reason-through-problems/ https://machinelearning.apple.com/research/illusion-of-thinking // Bill Gates says Spam will be gone by 2006 https://www.nytimes.com/2004/01/26/IHT-gates-sees-end-to-spam-in-2-years.html // Ships can't be hacked and driven into things: https://www.youtube.com/watch?v=5kkP-3d_8ac // Anthropic Essay conflating "Recursive Self-Improvement" with "8x more lines of code" https://www.anthropic.com/institute/recursive-self-improvement// Ties between AI Doomers, Rationalists and AI Employees https://x.com/kevinnbass/status/2099621874279817638 00:00 - Real AI Harms vs. Extinction Hysteria00:40 - Casey Hart & Carl Brown: "Prominent X-Risk Denialist"03:04 - Deconstructing the Doom Debate04:47 - Setting the Terms: What Does "Doom" Actually Mean?07:18 - Bayesian Thinking & The Base Rate Fallacy12:06 - Where Do p(doom) Percentages Come From?15:01 - The Core Doom Argument Broken Down20:47 - Premise 1: Is Intelligence Sufficient for Domination?22:36 - Did AI "Almost Start a War"? Misleading Headlines26:06 - Hanlon's Razor & Intentionality in LLMs29:25 - Premise 2: Limits to Superhuman Intelligence34:07 - Premises 3–5: Desires, Alignment, and Losing a Conflict49:33 - The Math of Multiplying Uncertain Premises53:31 - The Argument from Exponential Progress57:46 - Reasoning Models, Wrappers, and Stacking Advances59:40 - Benchmark Contamination & Overstated Gains01:04:03 - Real-World Coding Regressions (Fighting with Models)01:08:56 - Appeal to Expert Anxiety & Regulatory Moats01:16:07 - Assessing Authority vs. Vested Commercial Interests01:22:45 - The Argument from Unpreparedness & Cybersecurity Debt01:29:13 - The "Big Impacts" Fallacy & Infinite Disutility01:34:42 - What Does Recursive Self-Improvement (RSI) Actually Mean?01:41:08 - Anthropic's RSI Evidence: Checking in Lines of Code01:46:00 - Asymptotes, Data Bottlenecks, and Real-World Friction01:53:34 - Why Focusing on Fiction Harms Real-World Victims01:55:25 - Closing Thoughts: Stay Calm and Focus on Real Problems#ArtificialIntelligence #AISafety #TechPodcast #Philosophy #XRisk #MachineLearning #ExistentialRisk #AIEthics