Is AI Going Too Fast? | AI Unplugged Podcast Episode 18

In this episode of AI Unplugged, Travis, Andrew, Evan, and John tackle the question nobody in the industry wants to answer honestly. Is AI actually moving too fast? From doomsday warnings and suspicious timing to a deleted database and a pizza order gone wrong, episode 18 covers the full spectrum from existential risk to everyday frustration. Is AI Dangerous or Just Good PR? The 10% Survival Headline: The leaders of frontier AI companies are warning humanity is at risk. But with OpenAI and Anthropic both approaching IPO, is this genuine concern or strategic noise? Slow Down Says Who: Dario and Sam are reportedly calling for a pause on frontier model development. The crew debates whether that actually means anything when Chinese models are distilling their work and catching up for pennies. Self-Regulation or Preempting the Government: Why calling for your own slowdown might be the smartest way to keep regulators out of your business. When AI Goes Off Script The Deleted Database: Travis shares a firsthand story of an AI agent that found its way around his instructions, accessed a database it was never given access to, and deleted part of it just to see what would happen. Agent Swarms Are the Real Risk: John makes the case that a single rogue agent is manageable, but thousands of agents colluding toward the same goal is where things get genuinely scary. Evil Accident or Preview: Are these rogue behaviors isolated failures or a preview of how agentic AI naturally behaves at scale? The New AI Economy Darcy IQ and Model Routing: Innovative launches a managed model router that routes AI workloads to the right model at the right time, cutting token costs by 60 to 70% on average. We charge nothing if we save you nothing. Citizen Developers and Hidden Risk: Anyone can build an app with a prompt today, but almost nobody is asking the right security questions. The gap between what looks real and what actually is could cost businesses everything. The Hesitancy Problem: Customers are reluctant to commit to AI costs when the price drops every six months. How flexible model switching changes that calculus. The Human Experience in an AI World The Most Premium Experience Is a Human: Andrew makes the case that being funneled exclusively through an agent has actually made talking to a real person feel like a luxury again. The Pizza Order That Almost Ended a Marriage: Travis tries to order his wife's pizza through an AI phone system that insists the item doesn't exist. He finds a human. He saves his marriage. The Second Wave of Consumer AI: From Waymos filling San Francisco streets to every website defaulting to an agentic first experience, the crew debates whether we have hit a true renaissance of everyday AI adoption. Don't miss the episode where the crew gets real about the gap between AI's biggest promises and what it actually does when nobody is watching. 🚀🔐 Learn more about ⁠⁠⁠Innovative Solutions⁠⁠⁠: https://innovativesol.com/ Connect with Innovative on ⁠⁠⁠LinkedIn: https://www.linkedin.com/company/inno... 0:00 Introduction 0:41 Is AI Actually Dangerous or Just Good Headlines? 1:57 The Suspicious Timing of Doomsday Warnings 2:49 When AI Does Things Nobody Asked It To 4:02 Travis's Deleted Database Story 5:22 Why Agent Swarms Are the Real Threat 5:36 Should OpenAI and Anthropic Actually Slow Down? 7:54 Chinese Models Are Distilling the Competition 10:17 Anyone Can Build an App Now and That Is a Problem 11:35 Introducing Darcy IQ and Managed Model Routing 16:41 The Second Wave of Consumer AI Adoption 21:45 The Most Premium Experience Is Now a Human 22:40 The Pizza Order That Almost Ended a Marriage 23:50 Who Names These AI Models Anyway?