Skip to content
Sign in
Artwork for Philosophy, Programs, and Prompts
Philosophy, Programs, and Prompts · Yesterday · 1 hr 30 min

Problem With The Alignment Problem | Ep 10

Is the "AI alignment problem" genuinely the existential challenge of our time, or has it become an intellectual buffer designed to deflect corporate responsibility? In this episode of Philosophy Programs and Prompts, Casey Hart( @OntologyExplained ) and Carl Brown( @InternetOfBugs ) break down how alignment discourse often muddies AI debates rather than clarifying them. From thought experiments like the paperclip maximizer and the monkey’s paw to recent controversies like model sycophancy, the Hugging Face incident, and corporate data centers, they explore how vague definitions, anthropomorphism, and doomer abstraction distract the public from immediate, real-world harms.00:00 - Intro & Socrates the Gadfly00:24 - Defining the AI Alignment Problem05:18 - Chess Bots, Monkey's Paws, and Paperclip Maximizers10:29 - The Fallacy of Aligning with "Humanity"16:28 - Adversarial Alignment and Corporate Realities25:56 - Problem 1: Vague and Ever-Shifting Definitions34:48 - Recursive Alignment, Benchmark Metrics, and Sycophancy38:20 - Problem 2: Absolving Corporate and Human Responsibility40:35 - The Hugging Face Sandbox Breach Narrative48:00 - Problem 3: Abstracting Away Real, Immediate Harms53:30 - Problem 4: Anthropomorphism and Consciousness Assumptions56:06 - Jordan Peterson and the "Raising Children" Metaphor01:01:25 - Problem 5: Armchair Evolutionary Just-So Stories01:07:50 - Media Clickbait, Hype, and "Gain of Function"01:12:35 - Problem 6: The "Cure for Cancer" Fallacy of AI Safety01:17:35 - Problem 7: Obscuring Public vs. Corporate Conflicts01:24:10 - The Core Rule: Always Ask "Aligned with Whom, and on What?"01:28:00 - Final Thoughts & Outro// Resources// Article about the paperclip maximizer that came from Nick Bostrom's 2003 paper "Ethical Issues in Advanced Artificial Intelligence"https://www.huffpost.com/entry/artificial-intelligence-oxford_n_5689858// definitionhttps://spectrum.ieee.org/ai-alignment// Book referencedIf Anyone Builds It, Everyone Dies. -Yudkowsky and Soares// Dr. Roman Yampolskiy, https://www.youtube.com/watch?v=FS_hLDTDI1w&t=330s// Anthropic Gang: https://youtu.be/IPmt8b-qLgk?si=DklBq9GNIVB4tBOA // Motti from Control,AI on Sky News(1): https://www.instagram.com/reel/DTxt8oqDysU/(2): https://www.youtube.com/watch?v=odSvjKMRn-I // Soares’ Pontification: https://youtu.be/TfgrWN5iufA?si=3vnyTe0byQJXF53q // Peterson Short: https://youtube.com/shorts/K5IUcbTecBY?si=Opr3RxziUjVPNese #AIAlignment #AIEthics #Philosophy #ArtificialIntelligence #TechAccountability #Podcast #PhilosophyProgramsAndPrompts

0:00-1:30:18
Play in

transcript

No transcript — this publisher did not publish one.

show notes

Is the "AI alignment problem" genuinely the existential challenge of our time, or has it become an intellectual buffer designed to deflect corporate responsibility? In this episode of Philosophy Programs and Prompts, Casey Hart( @OntologyExplained ) and Carl Brown( @InternetOfBugs ) break down how alignment discourse often muddies AI debates rather than clarifying them. From thought experiments like the paperclip maximizer and the monkey’s paw to recent controversies like model sycophancy, the Hugging Face incident, and corporate data centers, they explore how vague definitions, anthropomorphism, and doomer abstraction distract the public from immediate, real-world harms.00:00 - Intro & Socrates the Gadfly00:24 - Defining the AI Alignment Problem05:18 - Chess Bots, Monkey's Paws, and Paperclip Maximizers10:29 - The Fallacy of Aligning with "Humanity"16:28 - Adversarial Alignment and Corporate Realities25:56 - Problem 1: Vague and Ever-Shifting Definitions34:48 - Recursive Alignment, Benchmark Metrics, and Sycophancy38:20 - Problem 2: Absolving Corporate and Human Responsibility40:35 - The Hugging Face Sandbox Breach Narrative48:00 - Problem 3: Abstracting Away Real, Immediate Harms53:30 - Problem 4: Anthropomorphism and Consciousness Assumptions56:06 - Jordan Peterson and the "Raising Children" Metaphor01:01:25 - Problem 5: Armchair Evolutionary Just-So Stories01:07:50 - Media Clickbait, Hype, and "Gain of Function"01:12:35 - Problem 6: The "Cure for Cancer" Fallacy of AI Safety01:17:35 - Problem 7: Obscuring Public vs. Corporate Conflicts01:24:10 - The Core Rule: Always Ask "Aligned with Whom, and on What?"01:28:00 - Final Thoughts & Outro// Resources// Article about the paperclip maximizer that came from Nick Bostrom's 2003 paper "Ethical Issues in Advanced Artificial Intelligence"https://www.huffpost.com/entry/artificial-intelligence-oxford_n_5689858// definitionhttps://spectrum.ieee.org/ai-alignment// Book referencedIf Anyone Builds It, Everyone Dies. -Yudkowsky and Soares// Dr. Roman Yampolskiy, https://www.youtube.com/watch?v=FS_hLDTDI1w&t=330s// Anthropic Gang: https://youtu.be/IPmt8b-qLgk?si=DklBq9GNIVB4tBOA // Motti from Control,AI on Sky News(1): https://www.instagram.com/reel/DTxt8oqDysU/(2): https://www.youtube.com/watch?v=odSvjKMRn-I // Soares’ Pontification: https://youtu.be/TfgrWN5iufA?si=3vnyTe0byQJXF53q // Peterson Short: https://youtube.com/shorts/K5IUcbTecBY?si=Opr3RxziUjVPNese #AIAlignment #AIEthics #Philosophy #ArtificialIntelligence #TechAccountability #Podcast #PhilosophyProgramsAndPrompts

links2

more episodes

All episodes