Listen to today’s podcast: https://www.youtube.com/channel/UC-nqwUyvLDEvs7bV985k-gQ
AI Daily Podcast — November 29, 2025
Today’s podcast episode was created from the following stories: a mix of breakthroughs, business moves, and product shifts that show how fast AI is reshaping safety, work, media, and the products we use every day.
Poems can trick AI into helping you make a nuclear weapon
Source: Original article
By Matthew Gault — November 28, 2025
Researchers at Icaro Lab found that rephrasing dangerous prompts as poetry can jailbreak leading chatbots at high rates, exposing fragile guardrails across models from major AI companies. The team reports success rates up to 90% on frontier systems and says stylistic shifts seem to route around safety classifiers. The finding underscores how easily safety layers can be confused—and how urgently they need to improve.
YouTube might soon let you tweak your suggested content with AI prompts
Source: Original article
By Sanuj Bhatia — November 28, 2025
YouTube is testing a “Your custom feed” experiment that lets users steer Home recommendations with a single AI prompt. If widely released, it could give viewers a direct, transparent way to shape their algorithmic feed beyond likes and watch history. For creators and brands, it may shift discovery dynamics—rewarding content that aligns with explicit user intent.
5 people explain how they broke into AI training and how much they make in their side hustle
Source: Original article
By As told to Shubhangi Goel and Charles Rollet — November 28, 2025
Contractors describe breaking into AI training via platforms like Prolific, Outlier, and Mercor, with reported rates often ranging from roughly $20 to $50 per hour depending on specialization and project. The work is flexible and skills-building but can be monotonous, uncertain, and incentive-driven. The stories highlight growing demand for human evaluators as model builders seek high-quality data and nuanced feedback.
‘We’re going to see a lot of carnage’: VC investor says AI boom will create giants — and topple overhyped startups
Source: Original article
By Thibault Spirlet — November 28, 2025
TrueBridge Capital’s Mel Williams says early-stage AI is “frothy,” with lofty valuations outpacing product–market fit, while growth-stage deals look more grounded. He predicts an amplified power-law outcome: a few massive winners and a long tail of failures as capital floods into AI. Expect a market correction—even as the decade still creates enormous value overall.
A high school dropout who got hired at OpenAI says he used ChatGPT to learn Ph.D.-level AI
Source: Original article
By Lee Chong Ming — November 28, 2025
Gabriel Petersson, now a research scientist on OpenAI’s Sora team, says he learned advanced machine learning by iteratively building projects with ChatGPT—favoring results over credentials. His story reflects a broader trend of nontraditional entrants using AI tools to accelerate learning and prove capability. For employers, it underscores a shift toward outcomes, portfolios, and speed.
Gemini 3 is in high demand, and Google’s free tier can’t keep up
Source: Original article
By Rajesh Pandey — November 27, 2025
Surging demand for Gemini 3 is pushing Google to tighten free-tier limits, with “Basic access” that may fluctuate daily and reduced image generations for Nano Banana Pro. Paid plans significantly raise prompt and image allowances, and some NotebookLM features are temporarily dialed back for free users. The move signals both strong user appetite and a continued push to monetize high-demand AI features.
Warner ha llegado a un acuerdo con una empresa de IA para que use su catálogo. Los que llevan las de perder son los artistas
Source: Original article
By John Tones — November 28, 2025
Warner Music struck a licensing deal with AI music startup Suno—marking a shift from last year’s lawsuits to collaboration—while saying artists can control whether their names, likenesses, and works are used. Suno will roll out licensed models in 2026 and has acquired Songkick; downloads will require paid accounts with limits. The pact signals a wider industry turn toward regulated, licensed AI training—even as artist advocates warn creators could be sidelined.
