Intelligent Machines Podcast Summary Episode IM 814: Chesterton's Fence - Cory Doctorow, Fake Meta Benchmarks, AI Therapy
Release Date: April 10, 2025
Hosts: Leo Laporte & Paris Martineau
Guest: Cory Doctorow

1. Introduction

In Episode 814 of Intelligent Machines, host Leo Laporte welcomes co-host Paris Martineau and special guest Cory Doctorow, a renowned sci-fi author, activist, and EFF consultant. The episode delves into a range of pressing AI topics, including AI art, benchmark manipulation by Meta, and the implications of AI in legal and healthcare sectors.

2. Book Discussion: Cory Doctorow's "Picks and Shovels"

Cory Doctorow introduces his latest novel, Picks and Shovels, the first in the Marty Hench Forensic Accountant series. Set in the early 1980s amidst the PC boom, the story follows Marty Hench as he navigates a world of DRM-locked computers and pyramid schemes, ultimately joining a rebellion against oppressive tech corporations.

"It's set in the early 1980s, which were, you know, the heroic era of the PC..." – Cory Doctorow [03:54]

3. AI Skepticism and Activism

Doctorow expresses his cautious stance on AI, identifying himself as a "committed AI skeptic." He emphasizes the importance of understanding AI's limitations and its potential to exacerbate existing societal issues, such as surveillance capitalism and the right to repair movements.

"I am a AI skeptic. And not just an AI skeptic, but a violent AI skeptic." – Cory Doctorow [09:38]

4. AI Art: Theory and Critique

A significant portion of the discussion centers on AI-generated art. Doctorow argues that AI art lacks the "big, numinous, irreducible feeling" that human artists embed in their work. He employs information theory to suggest that AI-generated pieces are merely probabilistic outputs without genuine intent or emotional depth.

"AI art feels at best eerie, where eeriness is defined as the seeming of intent without an intender." – Cory Doctorow [19:17]

Paris Martineau counters by highlighting that traditional art also involves substantial viewer interpretation, while Leo Laporte remains open to AI's potential to add novel connections.

"Artists have an authorial intent that AI lacks, making AI art feel soulless." – Paris Martineau [20:11]

5. AI Benchmarks and Meta's Llama 4

The episode addresses Meta's recent controversy over its AI model, Llama 4. Doctorow criticizes Meta for manipulating benchmarks to appear superior, undermining the integrity of AI evaluations.

"Meta fudged the benchmarks with their release of Llama 4. This confirms, by the way Corey, what you've been saying: To appear better, Meta manipulated the tests." – Leo Laporte [53:31]

Doctorow references Goodhart's Law, noting that when benchmarks become targets, their validity diminishes, leading to misleading representations of AI capabilities.

"Any benchmark becomes a target and then ceases to be useful as a benchmark." – Cory Doctorow [54:29]

6. AI in Radiology and Healthcare

Doctorow discusses the integration of AI in radiology, emphasizing that while AI can assist in detecting tumors, it often increases the workload for radiologists through false positives. He introduces the concept of the "moral crumple zone," where human professionals bear the blame for AI's errors.

"AI-based nonsense...movies accuracy...radiologists act as accountability sinks." – Cory Doctorow [48:25]

Leo Laporte highlights a Washington Post article indicating that AI in radiology has not killed the profession but is modifying it, requiring radiologists to verify AI findings, thereby not reducing their workload.

"AI hasn't killed radiology, but it is changing it." – Leo Laporte [48:25]

7. AI and Material Science

The conversation shifts to AI's role in material discovery. Doctorow critiques DeepMind's claim of accelerating material science by 800 years, revealing that a significant majority of AI-generated materials were deemed unusable.

"Most of them required that they would only exist at absolute zero." – Cory Doctorow [36:52]

Laporte poses that successful AI-driven material discoveries, like curing cancer, would validate AI's practical utility, contrasting DeepMind's largely unsuccessful outputs.

"If you came up with a cure for cancer, that would validate the work." – Leo Laporte [37:18]

8. Reshoring Manufacturing and AI's Impact on Jobs

Doctorow explores the challenges of reshoring manufacturing in the AI era, using the example of advanced microprocessor production. He underscores the improbability of attracting skilled labor to high-precision, low-paying jobs in the U.S., despite government initiatives like the CHIPS Act.

"It's not going to happen tomorrow. It will require complex, long-run phenomena to play out." – Cory Doctorow [66:06]

Laporte agrees, noting the deep-seated complexities in reversing global manufacturing trends.

9. AI Therapy: Risks and Privacy Concerns

The episode touches on the burgeoning trend of AI-based therapy. Doctorow warns of the dangers associated with entrusting sensitive mental health data to AI systems owned by companies with questionable data practices.

"This data is extremely compromising, and you have no basis for believing them." – Cory Doctorow [107:50]

Laporte references an NPR article suggesting the popularity of AI therapy, raising concerns about data privacy and the efficacy of such services.

10. Legal Battles: Section 230 and EFF

Doctorow updates on the EFF's lawsuit against Meta (referred to as "doge") over privacy violations related to data sharing with federal agencies. Despite the court's initial dismissal attempts, the case advances under the Administrative Procedures Act, signaling significant legal battles ahead concerning AI and data privacy.

"We are suing doge on behalf of two unions representing federal workers under the Privacy Act." – Cory Doctorow [58:41]

Laporte emphasizes the potential repercussions of Section 230 reforms, which could devastate smaller platforms by exposing them to lawsuits they cannot afford.

"Section 230's repeal would force small platforms like ours to shut down due to insurmountable legal costs." – Leo Laporte [74:03]

11. Conclusion: Reflections on AI's Future

The podcast concludes with a reflection on AI's dual-edged impact. While acknowledging AI's potential to innovate and solve complex problems, Doctorow and Laporte stress the need for cautious, regulated advancement to prevent societal harms. They advocate for viewing AI as a tool to augment human capabilities rather than replace them.

"AI companies are focused on monetizing current technologies rather than addressing speculative future threats." – Cory Doctorow [51:56]

"The ultimate criterion for AI's value is its ability to add meaningful information beyond mere computation." – Leo Laporte [24:19]

Notable Quotes:

• "AI art feels at best eerie, with the seeming of intent without an intender." – Cory Doctorow [19:17]
• "Heidi Goldman... accidentally accused someone using AI-generated information." – Cory Doctorow [48:09]
• "Meta fudged the benchmarks with their release of Llama 4 to appear superior." – Leo Laporte [53:31]
• "AI in radiology creates a moral crumple zone, increasing radiologists' workloads." – Cory Doctorow [48:25]
• "Section 230's repeal would force small platforms like ours to shut down." – Leo Laporte [74:03]

This episode of Intelligent Machines offers a critical examination of AI's evolving role in various sectors, highlighting both its transformative potential and the ethical, legal, and practical challenges it poses. Cory Doctorow's insights provide a grounded perspective on the promises and pitfalls of intelligent machines in the 21st century.