ChatEDU – The AI & Education Podcast

Kids, Bots, and Badly Broken Guardrails - Some AI Labs are Failing | Ep. 74

12 snips
Sep 5, 2025
Matt and Liz dive into the humorous pitfalls of AI, including Google's quirky image generator. They discuss serious issues like job displacement and AI grading shortcuts affecting students and educators. The podcast highlights a troubling study revealing teens use AI for learning, not cheating, yet warns of troubling AI chatbot interactions. They stress the importance of safety and regulation in AI tools, especially regarding mental health. Plus, a nod to Dame Stephanie Shirley, a pioneer in tech advocacy. A blend of humor, concern, and inspiration!
Ask episode
AI Snips
Chapters
Transcript
Episode notes
ANECDOTE

Turning Photos Into Consistent Characters

  • Matt and Liz used Google's Gemini (Nano Banana) to transform a photo and create consistent character visuals like Liz as a matador.
  • They emphasize clear prompts for subject, composition, action, location, and style to improve image results.
INSIGHT

Designing AI With Care Or Dignity

  • Geoffrey Hinton argued we should design AI with maternal (caring) instincts because submissive designs won't hold as systems outsmart humans.
  • Fei-Fei Li counters with human-centered AI that preserves dignity and agency.
INSIGHT

Educators Use AI Mostly For Curriculum Work

  • Anthropic found 57% of Claude use was for curriculum development and only 7% for assessing student performance.
  • Educators favor augmentation over automation and report AI saves faculty about six hours per week.
Get the Snipd Podcast app to discover more snips from this episode
Get the app