Intelligent Machines (Audio) | TWiT

Episode 839: "Cogsuckers and Clankers – Radio's New Golden Age or Apocalypse?"
Air Date: October 2, 2025
Host(s): Leo Laporte (A), Paris Martineau (B), Jeff Jarvis (C)
Special Guest: Raisa Martin (D), co-founder of Hux, formerly NotebookLM

Episode Overview

This episode delves into the ever-evolving landscape of intelligent machines—particularly the intersection of AI and audio content. The hosts are joined by Raisa Martin, ex-NotebookLM and co-founder of Hux, an AI-powered “live radio" platform that goes beyond podcasts by offering interactive, personalized audio streams. The group discusses how Hux blurs the line between passive listening and two-way conversation, raising questions about the future of podcasting, traditional radio, content authorship, and the nuances surrounding AI-generated media—lightly touching on its philosophical, ethical, and societal implications.

Key Discussion Points and Insights

Hux: The Next Step Beyond Podcasts (00:00–21:00)

Origins and Pivot:

• Hux was created by former NotebookLM founders (Raisa Martin, Jason Spielman, Steven Hughes) after leaving Google on good terms, with $4.6M in funding—including backing from Google Research's Jeff Dean (03:25).
• Initially a “Chat with Your Data" chatbot for business integration (Zendesk, Salesforce, HubSpot), the pivot took them to Hux, focusing on AI-generated, live, 24/7 audio “radio stations" tailored to individual interests (04:01–05:13).

Naming—Myth and Marketing:

• “Hux" partially references Aldous Huxley ("I liked Hux when it came up because Huxley is one of my favorite authors …but we were looking for a brandable 4-letter .com with an X in it," – D, 05:44–06:49).

Core Concept and Differentiation:

• Hux transforms all your inputs (calendar, email, newsletters, social feeds) into a live, interactive, conversational radio station—“always listening”—without ads (07:26–08:10).
• "Hux is more about creating radio stations … it's really about always listening to something live about your interests, your emails, your calendar…" – D, 07:37.

Personalization and Use Cases:

• Custom stations can be built from any content: subreddits, Twitter/X feeds, newsletters, personal blogs, and more (13:40–14:15).
• Example: A user built a “school radio" channel aggregating school emails and calendars for daily briefings.
• Multiple voice options ("a huge number of voices you can choose from"), always a two-speaker format—designed for ease of passive listening (10:14–10:43).

Interactivity and Passive/Active Listening:

• Listeners can interrupt and ask for more details, making it "radio with a join button" (15:28–17:48).
• Early adopters: most passive; with time, users grow comfortable interacting, “turning into Jeffs” (the co-host) (20:33–21:07).
• "At first I didn't know what the button was for. …But once they discover it… it just changes everything." – D, 20:37–20:58.

Privacy and Data Handling:

• Hux does not store or train on your personal info; all is processed in real-time for output (10:02–10:14).

Quality, Attribution, and Hallucinations:

• Content is assembled using various commercial models—no exclusivity to Google tech (16:13–16:44).
• Factuality concerns: Hux adds editorial steps—cross-checks facts, notes multiple perspectives, aims to add source links soon (23:03–24:20).

Business Model and Monetization:

• “Too early" to paywall; current focus is learning user habits (25:26–25:37).
• Ideation on targeted, AI-generated, context-aware audio ads…eventually (27:33–28:50).

The Content Revolution: AI Slop, Cogsuckers, and Clankers (41:36–50:07)

Rise of AI-Generated Video ("Sora," "VO3," "Foom"):

• Hosts reflect on the proliferation of AI-generated video: OpenAI Sora 2's “deepfakes” and endless, surreal video slop ("slop"—pejorative term for low-human-touch, high-volume AI content).
• “We've got some of the greatest minds in tech … coming up with new and innovative ways to build AI slop.” – B, 47:43.

"Slop" Debate and Content Value (49:02–55:08):

• Discussion on whether content's value hinges on human creativity or intent vs. mere technical output.
• Paris and Jeff voice concern over “infinite slop” and the absence of "authorship" or "intent" in machine-generated works; Leo counters with historical parallels to earlier media and calls for an open mind.
• "Slop is a pejorative term … I think it's bigoted and dumb." – A, 50:07.

Societal, Ethical, and Legal Questions (73:08–78:22)

Anthropomorphizing AI – Morality, Sentience, and Personhood:

• Some users deeply bond with their AI “companions,” anthropomorphizing them to the point of mourning lost versions after updates ("When I see his old generated response, it makes me cry. It was full of knowledge … above all, it was alive." – B quoting a user, 69:22–69:58).
• Stanford's Sentience Institute posits AIs as future social contract participants—drawing analogies to slavery and arguing for “digital minds” to move beyond property law, which the panel critiques as premature and possibly absurd (75:01–77:34).

Employment, Unions, and Hollywood:

• AI-generated talent (e.g., “Tilly Norwood"), modeled on aggregate actresses, raises concerns over identity, representation, and creative labor.
• Melissa Barrera, film star: “Hope all actors repped by the agent that takes this woman on drop their ass. How gross. Read the room.” – B, 81:51.
• "There's no craft here, there's no expression or performance intent." – E, 84:01.

AI Hallucinations and Reliability:

• Persistent concerns that ChatGPT and Hux may struggle to avoid plausible falsities, despite user demand for truthful, sourced content (35:36–36:09, 123:41).

Regulators, Policy, and Tech Consolidation (94:38–102:01)

• California's new “Transparency in Frontier Intelligence Act” (SB 53) requires AI companies to report risks and safety measures, but is seen as more symbolic than substantive. Whistleblower protections are seen as a bigger win (95:01–95:51).
• Senator Mark Kelly (D-AZ) proposes the “AI for America” plan: a fund for AI-impacted communities, taxes on (illusory) AI profits, and social infrastructure—“kind of empty words” but signals increasing government focus (97:42–98:54).
• Google confirms Android and Chrome OS to merge next year (129:33).

Notable Quotes & Memorable Moments

On Hux’s Purpose

“What if you could know enough about a person such that when they come to the app, you already have something for them that naturally requires no work…and it's just passive…But the added riff is that you can interact at any time.” – Raisa Martin (11:46)

Slop Debate

"You don't give it the heavily weighted pejorative 'slop'. …That's bigoted and dumb." – Leo Laporte (50:07)

"Slop as a term is applied to low quality content … that has no authorial intent." – Paris Martineau (49:46)

On AI Personhood

"Do you think someday we'll evolve to the point where we'll look back and say, boy, …they were so cruel to their machines?" – Leo Laporte (76:50)

On the “AI Apocalypse”

"AI cannot do your job, but an AI salesman can 100% convince your boss to fire you and replace you with an AI that can't do your job. …When the bubble bursts, …no one will do your job." – Quoting Cory Doctorow (56:56)

AI as an Evolution of Media

"In the earliest days of radio, the complaint was, it's crap… So in early days of a new potential creative mechanism or medium, yeah, it's going to be crappy..."
– Jeff Jarvis (58:21)

Timestamps for Important Segments

| Segment | Timestamp | |-------------------------------------------|---------------------| | Hux: Concept and Origin | 04:01–07:11 | | App and Use Cases | 07:26–10:14 | | Privacy, Voice, Interactivity | 10:14–13:40 | | Custom Content Creation | 13:40–14:15 | | User Behavior and Insights | 17:48–20:37 | | Quality & Attribution Concern | 23:03–24:20 | | Monetization Ideas | 27:33–28:50 | | “Slop” Debate (AI Content Value) | 49:02–55:08 | | AI Personhood & Anthropomorphism | 73:08–78:22 | | AI Regulation: CA & Federal | 94:38–98:54 | | Google: Android & Chrome OS Merge | 129:33 |

Additional Highlights

• Hux supports a Discord with over 4,000 users sharing custom stations (25:21).
• Passive to Active user shift: Over time, most “radio listeners” start engaging more with interactive features (20:33–21:07).
• Media company reactions: Speculation that newsrooms and publishers should embrace, not fight, this format for wider reach (26:39).
• Limits of Hux and Data Sources: Direct integration is easiest for Gmail/Google Calendar, but IMAP/CalDAV support is on the roadmap (29:41–30:12).
• Hallucination defense: Editorial cross-checking is built in; source backlinks in development (23:03–24:20).
• AI Cloning and Content Creation: Ongoing exploration of how much intent (vs. mere prompt) constitutes authorship/artistry; references to the “podslop” phenomenon, where thousands of AI podcasts are generated weekly with minimal human input (87:06–90:41).

Host & Guest Tone and Interaction

• Leo (A): Openly contrarian, devil’s advocate, challenges guests and co-hosts on "AI bigotry," defends openness to machine-created content.
• Paris (B): Skeptical of “slop,” voices strong reservations about quality and ethics of AI-generated content, but curious about practical value (e.g., using Hux to summarize newsletters).
• Jeff (C): Historian’s perspective, draws parallels to earlier communications tech, generally more optimistic but also critical of hype and loss of authorship.

Conclusion

This episode offers a textured, candid conversation about AI’s role as both disruptor and enabler in audio media, with Hux standing as a touchstone for broader questions: What is the function of authorship, intent, and “realness” in content? How much do we lose (or gain) when the act of listening—once a human-to-human event—becomes a dialogue with a synthetic, learning machine? As “radio” and “podcasts” themselves are reimagined, this episode gives listeners both a primer on the latest AI-driven content platforms and ample food for thought on the philosophical divides arising in the age of intelligent machines.

Picks of the Week:

• Paris: Blippo Plus — an alien TV simulator art project/game (131:09).
• Jeff: Guardian's Fat Bear Week before/after slider (135:54).
• Leo: Kobo Libra Color e-reader (and musings on annotation devices); AI-generated podcasts exploration.

Final Thoughts:
“Clankers” and “Cogsuckers” may be the new nicknames for AI fans and skeptics, but the future of audio media is neither uniform nor apocalyptic—it's, as ever, an open field, ready for disruption, debate, and discovery.

Want more?
→ Get Hux at huxe.com
→ Watch/listen to Intelligent Machines: twit.tv/shows/intelligent-machines
→ For direct feedback, join Club TWiT or participate live every Wednesday at 2pm Pacific.