The AIs are officially out of control

Episode Summary

Executive Summary: The episode centers on how AI is reshaping the web, with a deep dive into Reddit’s data licensing deal with Google and the broader economics of training AI on human-generated content. The hosts also dissect Gemini’s image-generation diversity failures and OpenAI’s recent gibberish bug, then move through lightning-round coverage of ad-tech consolidation, smart-home hardware, Apple’s sports/betting ambitions, quantum-safe messaging, and EV market softening.

Main Topics: Reddit and Google’s AI data deal (Priority: 5/5): The hosts debate the significance of Google paying Reddit $60 million annually to train AI on the “Reddit corpus,” arguing it could reshape incentives for web publishing, search, and future AI licensing deals. AI image generation and culture-war blowback (Priority: 5/5): They analyze Gemini’s racially diverse image outputs and the resulting backlash, framing it as an unavoidable collision between generative tools, historical prompts, and cultural expectations. AI reliability and system failures (Priority: 4/5): OpenAI’s temporary gibberish output is discussed as evidence that even creators do not fully understand or control these systems, reinforcing skepticism about handing culture-making to AI. Ad-tech and retail media expansion (Priority: 4/5): Walmart’s acquisition of Vizio is presented as a move to expand connected TV advertising, reflecting how retailers are building ad networks to compete with Google, Meta, and Amazon. Smart-home and consumer hardware frustrations (Priority: 3/5): The discussion ranges from Samsung’s AuraCast rollout and the Echo Hub to a long-running rant about the Frame TV’s art mode and API issues, illustrating the tension between useful hardware and annoying software. Apple, sports, and gambling adjacency (Priority: 3/5): The hosts examine Apple Sports as a possible precursor to deeper sports-streaming or gambling integration, with skepticism about whether the app is merely a toy or a strategic wedge. EV market correction and product positioning (Priority: 3/5): Rivian layoffs, Ford price cuts, and broader EV demand slowdown are used to argue that the market is normalizing after pandemic-era hype, especially for expensive trucks and SUVs.

Key Arguments: Reddit has unusually valuable leverage because its user-generated discussions are critical for search and AI training, making the $60 million annual Google deal potentially too low or at least a benchmark-setting number. Google’s dependence on Reddit-like human conversation may intensify as AI-generated search answers degrade, pushing Google to prioritize real forum content more heavily. Gemini’s image issues show that generative AI cannot cleanly separate “art” from “bias” or “truth” when prompts involve historical people, identities, or sensitive categories. OpenAI’s gibberish episode highlights the fragility of generative models and how little even their makers understand about their internal behavior. Retailers like Walmart are turning TVs into ad inventory, suggesting connected TV is becoming a major battleground in advertising. Apple Sports looks like either a small utility app or an early move toward sports betting integration; either way, the hosts think Apple is closer to gambling than many assume. The EV boom exposed a mismatch between hype and actual demand, especially in the premium truck segment, and the next phase will likely reward mainstream pricing over novelty or scale promises. Smart-home ecosystems still depend too much on fragile apps and voice control; physical interfaces like the Echo Hub are more reliable than trying to force phones and assistants to do everything.

Data Points: Reddit–Google licensing deal: $60 million per year - Google will pay Reddit annually to train AI systems on Reddit data. Reddit annual revenue: ~$450 million per year - Used as rough context for why $60 million is material to Reddit. LinkedIn Ads offer: $250 spend + $250 credit - Ad read promotion for first campaign on LinkedIn Ads. Wealthfront cash account APY: 4% APY - Mentioned in sponsor read for uninvested cash. Vizio acquisition price: $2.3 billion - Walmart’s purchase of Vizio to expand retail media and connected TV advertising. Vizio/Walmart ad strategy: Connected TV advertising platform - Walmart explicitly framed the deal around ad-tech and Walmart Connect. Gemini image generation issue: Diverse Nazis / diverse historical figures - Examples used to illustrate image prompt failures and overcorrection. OpenAI failure duration: About 8 hours - ChatGPT produced gibberish during a temporary system issue. Apple Sports betting odds integration: On by default - Odds are shown in the Apple Sports app via DraftKings, though users can disable it. Rivian workforce cut: 10% - Rivian layoffs amid slowing EV demand and pricing pressure. Quantum-safe messaging: Post-quantum cryptography for iMessage - Apple’s new security update for US users. Rivian product pricing: Around $100,000 at top trims - Used to explain limited demand for premium electric trucks/SUVs.

Pivotal Quotes: "Today you've contributed to the corpus." — David Pierce: Joking about Reddit’s role as a data source for AI training and search. "This is the beginning. This is the end of the beginning of this phase." — David Pierce: Assessing the significance of the Reddit-Google licensing deal for the future of web content and AI. "It is a big bummer back in the wrong direction." — Alec Trance: Reacting to Google’s new Gemma naming direction and broader product-naming choices.

Implications: AI licensing, search quality, ad-tech, and consumer hardware are converging into a new internet model where human-generated data and attention are monetized more directly. The winners will likely be platforms with leverage, while users face more ads, more friction, and more AI-driven instability.