Doom Debate Prep | Ep 11
In this episode, Casey( @OntologyExplained ) and Carl( @InternetOfBugs ) prep for Casey's upcoming debate on ( @DoomDebates ) with Liron Shapira. They dive into what "P-Doom" actually means, break down the feasibility of recursive self-improvement (RSI), and evaluate whether recent breakthroughs—like AI-assisted Lean proofs for Navier–Stokes—actually point toward artificial superintelligence or just sophisticated, narrow mixtures of experts. Along the way, they analyze the incentives driving public AI panic, the rhetorical tactics of the rationalist community, and why focusing on speculative existential risk often distracts from tangible, real-world AI harms.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/// References and Resources:// Ed Dows explaining the recent high-profile Anthropic resignation as a PsyOphttps://eddowdbeyondthenarrative.substack.com/p/the-post-labor-day-ai-panic-psyopWatch chatbbots attempt to play chess in hilarious ways @GothamChess ( https://www.youtube.com/playlist?list=PLBRObSmbZluQwBIvxyiWfm_b495AnpzDn )// Mac app Carl mentioned that lets Gemma 4 26B run in 2GN of RAM by only loading the "expert" that's needed at the moment: https://github.com/drumih/turbo-fieldfare/More on decomposing LLMs at @chrishayuk // On the Apple Paper about how chatbots fail problems that look similar to but aren't benchmark problems because they follow the wrong patternshttps://arstechnica.com/ai/2024/10/llms-cant-perform-genuine-logical-reasoning-apple-researchers-suggest/https://arxiv.org/pdf/2410.0522900:00:00 - Cold Open: The Moron or Masochist Rule00:00:11 - Intro: Preserving Sanity Amid AI Hype00:00:46 - Prepping for Doom Debates with Liron Shapira00:03:09 - The Value (and Trap) of Public Debates00:07:37 - What Exactly is P-Doom? Defining the Stakes00:11:45 - Knowing the Audience: Preaching to the Converted vs. the 1%00:15:38 - Finding the Common Ground (and the Common Enemy)00:21:40 - Deconstructing *If Anyone Builds It, Everyone Dies*00:26:22 - The 2050 Extinction Baseline00:29:44 - Recursive Self-Improvement: Hype vs. Reality00:33:41 - AI Takeover vs. Bad Actors with Powerful Tools00:41:43 - The Bayesian Argument Against Doom00:44:13 - A Bug-Centered Epistemology00:47:50 - The Economic Engine: Capital, IPOs, and Regulatory Capture00:52:03 - True Believers and Singularity Evangelists00:56:01 - Manufactured Hype and Coordinated Resignations01:00:00 - Critique of Rationalist Culture and Intelligence Fetishization01:01:43 - The Orthogonality Thesis: Wants vs. Intelligence01:02:40 - Navier–Stokes, Lean Prover, and Mixture of Experts01:03:40 - Goodhart’s Law, P-Hacking, and Jagged Intelligence01:04:46 - Extraordinary Claims Require Extraordinary Evidence01:05:15 - Post-Show: Podcasting Logistics and Editing Legos#AIDoom #ArtificialIntelligence #AISafety #TechDebate #MachineLearning #PhilosophyProgramsAndPrompts #DoomDebates