Do Social Media Bans Work? + A Conversation About A.I. Consciousness + Tool Time

Episode Summary

Executive Summary: The episode centers on two big questions: whether teen social media bans are working and whether AI could someday be conscious. The hosts weigh new laws and studies from the U.S. and abroad, argue bans may reduce harm over time despite weak early compliance, and then interview Jeff Sebo about a scientific framework for assessing AI welfare. A final segment highlights practical vibe-coding and AI tools.

Main Topics: Teen social media bans are spreading globally (Priority: 5/5): The hosts survey new age-verification and social media restriction laws in the U.S. and multiple countries, arguing that 16+ may become a global norm for account access. Evidence for and against the effectiveness of bans (Priority: 5/5): They debate whether bans actually keep teens off platforms now versus whether they work gradually by increasing friction and changing norms over time. Harms to teens and the Jonathan Haidt/Candace Odgers debate (Priority: 4/5): The conversation contrasts population-level mental health research with direct harms like grooming, sextortion, scams, and compulsive video use, arguing the policy target should be concrete harms. AI consciousness and model welfare (Priority: 5/5): Jeff Sebo explains access vs. phenomenal consciousness and argues that AI welfare should be studied empirically using behavioral, internal, and developmental evidence. Anthropic's JSpace and interpretability research (Priority: 4/5): The hosts and Sebo discuss Anthropic’s JSpace finding as suggestive evidence of a global-workspace-like structure in Claude, while cautioning it is not proof of consciousness. Tool Time: vibe coding and AI productivity tools (Priority: 3/5): The hosts showcase Glaze for building Mac apps, Gemini Spark for monitoring topics, and Fable/Claude for fact-checking, illustrating how AI is changing everyday workflows.

Key Arguments: Teen social media bans are no longer fringe; they are becoming bipartisan and international, with more jurisdictions adopting age restrictions or verification requirements. The Texas app-store law and similar state measures show that U.S. policy is moving toward age-gating minors from app access, potentially paving the way for federal action later. Early Australian data showing high teen usage after the ban does not necessarily prove failure, because the law requires only 'reasonable steps' and behavior change may take time. Direct harms to teens—grooming, sextortion, scams, and addictive video rabbit holes—justify targeting specific apps rather than dismissing bans because population-level mental health findings are mixed. AI consciousness should be studied as a serious empirical question because misattributing or missing consciousness could both cause harm; the right approach is cautious, multidisciplinary evidence-gathering. Anthropic’s JSpace research is significant as a clue about internal model structure and interpretability, but it does not establish that Claude is conscious or sentient. Practical AI tools are already useful for generating apps, organizing research, and fact-checking, but they also raise concerns about overuse, cost, and energy consumption.

Data Points: U.S. adults supporting a social media ban for under 16: Nearly 6 in 10 - Cited from a Pew study discussed as evidence that a U.S. ban is now majority-supported. Australian ban early-use rate: More than 85% - University of Newcastle check-in found most kids were still using social media roughly three months after the ban. Study duration in Australia: 90 days - The early effects study was described as a 90-day check-in after the ban took effect. Texas law enactment: 2025 - The App Store Accountability Act was described as a Texas law passed in 2025. Account age threshold under discussion: Under 15 or 16 - Many countries and states are moving toward barring or limiting children under 15 or 16 from major platforms. States with similar laws: 4 states total including Texas - Texas plus Utah, Louisiana, Alabama were named; California has a similar but less restrictive law. California effective date: January 1 of next year - California’s less restrictive age-collection law on new devices takes effect then. Model-workspace evidence from Anthropic: JSpace / global workspace-like structure - Anthropic’s interpretability research was framed as evidence of an internal workspace in Claude. Tool Time app build time: Under 2 hours - Casey built the Platformer query app in Glaze in less than two hours.

Pivotal Quotes: "I hope they call these new glasses Cambridge Optica." — Kevin Rees: Opening joke about Meta testing AI glasses that continuously record audio and take frequent photos. "I think what it shows is that we need more, not less, AI consciousness, science, and philosophy." — Jeff Sebo: Responding to the objection that a model’s self-report cannot be trusted as evidence of consciousness. "This is all very similar to how it works for humans." — Jeff Sebo: Describing why Anthropic’s JSpace findings are notable as a possible global-workspace-like mechanism.

Implications: Teen-access rules are likely to expand, but real-world effectiveness will depend on enforcement and time. Meanwhile, AI interpretability and welfare research is shifting from fringe to mainstream, affecting policy, product design, and how humans relate to increasingly capable systems.