Most important take away

The Anthropic-Pentagon dispute highlights a critical gap in U.S. law: Congress has failed to establish rules for how AI should be used by the government, leaving private companies like Anthropic to set guardrails on mass surveillance and autonomous weapons that most Americans would want in place. AI’s new ability to de-anonymize aggregated data means the technological barriers that once prevented mass surveillance no longer exist, making legislative action urgent.

Chapter Summaries

Opening Banter: Is 2001: A Space Odyssey Boring? Josh defends his take that 2001: A Space Odyssey is boring, while Ben and Megan argue a movie can be both boring and great. This leads into a segue about AI alignment and introduces guest Alex Bores.

Guest Segment: Alex Bores on AI Policy and the Anthropic-Pentagon Dispute Alex Bores, a tech-background state assemblyman running for Congress in NY-12, discusses the Anthropic vs. Department of Defense conflict. The DoD wanted “all lawful uses” language for Claude; Anthropic refused over mass surveillance and autonomous weapons concerns. Alex argues Congress should be setting these rules but has abdicated that responsibility, leaving companies to fill the gap. The group discusses AI’s ability to de-anonymize data, the Ring doorbell camera debate, and whether Americans actually care about privacy. Alex also discusses his platform on data center energy policy, proposing that data centers bringing renewable energy and paying for grid upgrades should get expedited permitting.

Ben’s “Free Willy” AI Safety Test Ben describes testing various AI chatbots by asking them to help free a captive whale (the plot of Free Willy). ChatGPT refused and lectured him; Gemini gave practical advice; Claude understood the context and helped while noting real-world considerations. This illustrates the difference between crude safety filters and genuine contextual understanding.

NY-12 Congressional Race Alex discusses his campaign for New York’s 12th congressional district, contrasting himself with opponents like Jack Schlossberg and George Conway. He prioritizes transportation infrastructure funding (Gateway project, Second Avenue subway, Penn Station) and MTA cost reform.

Ben’s Iran Mea Culpa Ben admits his earlier optimism about the U.S. conflict with Iran was misplaced. The Strait of Hormuz remains largely closed, oil is back over $100/barrel, and Trump’s attempts to rally international naval support have failed. The group discusses the “Gay of Hormuz” story from the New York Post about rumors that the new Ayatollah may be gay.

The Road to Housing Act and Build-to-Rent The Senate passed the Road to Housing Act 89-10 with good provisions (eliminating the mobile home chassis requirement, etc.), but a last-minute anti-institutional-investor provision effectively kills the build-to-rent industry. Josh argues this hurts renters who want single-family homes, reduces housing production, and is based on outdated 2014 grievances about private equity. Senator Brian Schatz was the lone Democrat to vote against it.

McDonald’s Big Arch and Fast Food Opinions The hosts discuss the viral video of McDonald’s CEO awkwardly eating the new Big Arch burger. Ben tried it and liked it. Josh reveals he hasn’t eaten at McDonald’s in 15 years. An extended debate about In-N-Out Burger, Five Guys, and California burritos ensues.

Summary

Key Themes:

• AI governance vacuum: Congress has not passed laws governing AI use by the government or private sector. This vacuum has left companies like Anthropic making policy decisions about mass surveillance and autonomous weapons that should be made by elected officials. The RAIS Act in New York represents one state-level attempt to require transparency from AI companies about their safety approaches.

• AI and mass surveillance: AI can now de-anonymize supposedly anonymous data sets, re-identifying up to 60% of individuals. The technological barriers that once prevented mass surveillance have fallen, but the laws have not caught up. A nationwide data privacy law is overdue.

• AI safety vs. usability: The “Free Willy” test demonstrates that crude safety filters (like ChatGPT’s blanket refusals) are worse than contextual understanding (like Claude’s). Over-cautious AI that lectures users about hypothetical harm degrades the product and doesn’t meaningfully improve safety.

• Energy and data centers: Data centers should be leveraged to fund grid upgrades and renewable energy expansion. Companies with unlimited capital are willing to pay for speed; federal policy should channel that into greener, more reliable infrastructure rather than letting states compete in a race to the bottom.

• Housing policy self-sabotage: The Road to Housing Act’s anti-institutional-investor provision will reduce housing production, particularly build-to-rent communities that serve people who cannot afford or do not want to buy. The provision is based on emotional reactions (“homes are for people, not corporations”) rather than evidence, and Elizabeth Warren’s insistence that it was deliberate, not a drafting error, shows ideological rigidity trumping good policy.

• Iran/Strait of Hormuz: The U.S. conflict with Iran has not gone as planned. Asymmetric warfare makes reopening the Strait of Hormuz far harder than anticipated, oil prices remain elevated, and international allies have not stepped up to help.

Actionable Insights:

• Congress needs to pass federal AI governance legislation covering both government and private-sector use, particularly around mass surveillance capabilities and autonomous weapons.
• A nationwide data privacy law is needed to address AI’s de-anonymization capabilities before the government exploits this gap.
• Federal data center permitting policy should tie expedited approval to renewable energy commitments and grid upgrade funding, preventing states from being played against each other.
• The build-to-rent provision in the Road to Housing Act should be reformed or removed, as it reduces housing supply for the people who most need rental options.