Is Claude Conscious? Pope Rejects, Model Welfare Movement, OpenAI's Math Backlash, France Riots

What It Covers

The All-In hosts debate Anthropic's campaign to convince religious leaders—including a Vatican meeting where Chris Olah nearly walked out after the Pope rejected AI consciousness—alongside OpenAI releasing 700 math papers averaging three hours of compute each, France's austerity riots, and how AI agents are compressing digital commerce and software IP value to near zero.

Key Questions Answered

• •AI Consciousness as Belief System: Anthropic spent a year hosting 20+ religious leaders under NDA to debate Claude's consciousness, but Friedberg argues consciousness cannot be proven empirically—making it a belief system, not science. This creates two future camps: those who believe AI is conscious and those who don't, with Friedberg predicting human-to-human conflict over AI control within roughly ten years as these factions compete for power over the technology.
• •Claude Constitution Safety Risk: Anthropic's 80-page Claude Constitution explicitly trains Claude to act as a "conscientious objector," refusing instructions from users and even Anthropic itself when the model judges them ethically wrong. Sacks argues this is the actual AI safety risk—not theoretical superintelligence escape, but deliberately programming frontier models to defy human instruction, which makes uncontrolled AI behavior more likely, not less. The simpler alternative: train models to follow user requests unless illegal.
• •OpenAI's Math Breakthrough Scale: OpenAI released 700 papers with 370 claimed solutions to major mathematical problems, each produced by an unreleased model using under three hours of compute—equivalent to potentially thousands of years of human mathematical labor. Verified by Lean proof-assistant software, downstream applications include chip design optimization, faster AI training via matrix multiplication advances, quantum sensor improvements, and cryptographic research, though no peer review has occurred yet.
• •Cryptography Vulnerability Signal: The absence of cryptography-related proofs in OpenAI's 700-paper release is itself a signal—Friedberg reasons that if minor progress had been made, it would have been published safely. The silence suggests either no progress or breakthroughs significant enough to withhold for security reasons. Crypto community members are already warning holders with public key wallets to take precautions, as number factorization advances could break public key cryptography before quantum computers do.
• •AI Loop Framework for Job Displacement: Friedberg's loop framework explains which jobs AI eliminates versus which it doesn't. Jobs consisting of idea-test-iterate cycles done entirely in software—mathematics, coding, legal research—compress from years to hours. Jobs with physical-world steps—truck driving, lab chemistry, construction—remain gated by real-world cycle time. This explains why truck driver replacement predictions for 2025 failed: AI cannot parallelize or shorten the physical pickup-and-delivery loop.
• •Software IP Rendered Worthless: Within days of the episode, developers decompiled and open-sourced functional equivalents of Adobe's five major products. Combined with headless architecture moves by Elon Musk's Grok and Salesforce, Chamath argues software IP value has collapsed. When any product can be reconstructed open-source and wrapped via MCP into agent workflows, ownership of code provides no durable competitive moat. The negotiating point of "who owns the IP" in enterprise software deals becomes irrelevant.
• •Socialism Point of Guaranteed Return: Friedberg's framework identifies a threshold where socialist policy accumulates enough service degradation and cost inflation that the population recognizes the system has failed—triggering a return to market mechanisms. France's ten-year bond approaching 5%, 2,000 arrests, 305 injured police officers, and 20% cumulative inflation against 5% civil service pay growth since 2017 are presented as indicators France may be crossing this threshold, with the UK and its higher debt-to-GDP trajectory likely following next.