In this episode, Paul brings a playful twist to Talking to AI by challenging ChatGPT to a few classic games — first a card game experiment, then a round of the timeless "Animal, Mineral, or Vegetable." What starts as a simple idea quickly becomes an exploration of how AI handles rules, reasoning, and even the unexpected moments of human humor that make games so fun.

From logical deduction to quick wit, Paul tests just how well AI can adapt to spontaneous gameplay — and what it reveals about how we communicate, improvise, and connect with technology. Can a machine really "play" the way humans do, or does it just simulate the fun?

Tune in for a lighthearted but fascinating look at AI word games, problem-solving, and the surprising charm of a digital opponent who never runs out of patience.

In this short and insightful chat, Paul dives into the hidden corners of the ChatGPT interface to uncover how to get more done — faster. Learn how "Projects" keep conversations connected, "Libraries" store your go-to resources, and "GPTs" act as custom AIs for specific workflows.

If you've ever wondered how to make ChatGPT a true productivity partner, this episode's for you.

For show notes and more, visit www.talkingtoai.show

Can AI really hold its own in a debate? Paul challenges ChatGPT to a formal, timed discussion on whether cities should invest more in public parks or public transport. The result is part experiment, part entertainment — and a glimpse into how structure, logic, and a touch of humor shape human–AI dialogue.

Can ChatGPT really help you learn a new language? Paul puts it to the test — creating a live Norwegian lesson with AI. Together, they explore pronunciation, translation, and how ChatGPT adapts to feedback.

A thoughtful, funny, and surprisingly human look at how people and AI learn together.

Can ChatGPT help you sleep through a crying baby? Paul puts AI to the test — asking how to block baby noises while staying comfortable. The conversation unpacks the real science behind earplugs, earmuffs, and noise-cancelling tech, separating myths from reality.

A surprising and practical episode about sound, sleep, and the strange new ways we rely on AI.

Why does ChatGPT sometimes hesitate, evade, or sound too careful? In this live uncut episode, Paul explores the principles that shape AI's responses — from safety to uncertainty — revealing what really happens behind the curtain of intelligent conversation.

Learn how to phrase smarter prompts and understand the psychology of talking to machines.

In this episode, I step back to ask a big, foundational question: what exactly is science, and how do we judge the quality of evidence? Before diving into future conversations on scientific topics, I wanted to better understand the scientific process itself.

With the AI, I explore the hierarchy of studies—from gold-standard randomized controlled trials, to cohort and case-control studies, all the way down to case reports. Along the way, we look at real-world examples and unpack the strengths and weaknesses of each approach: why some studies are more trustworthy than others, how bias creeps in, and what to watch for when evaluating research.

It's a clear, practical guide to thinking scientifically—helping me (and hopefully you) become more confident in navigating evidence, studies, and claims in everyday life.

In this episode, I turn the tables and ask ChatGPT for help—not about philosophy or creativity this time, but about something far more practical: how to host this very podcast.

What begins as a technical question turns into an unexpectedly thoughtful conversation about control, freedom, and digital longevity. Together, we explore the trade-offs between convenience and ownership, and I press the AI on how to avoid getting locked into platforms that could one day vanish or change.

As I weigh options like Libsyn, Podbean, and others, ChatGPT patiently breaks down costs, migrations, and redirects, while I push for clarity on what really happens when you move your podcast. The discussion ends up revealing as much about trusting technology as it does about podcast hosting.

It's part tutorial, part philosophical reflection—and another reminder that sometimes, the most human thing you can do with AI is just… ask a good question.

Image created via grok

So the first time I chatted with chatgpt I taught it to hum. However its still not very good. I fear it now thinks it can hum buts having real trouble. It is trying to sing now but it is not very good.

Second time round s is a funny and sometimes frustrating experiment in pushing its limits. I ask it to hum single notes, copy simple tunes like Twinkle, Twinkle, Little Star, and sustain long, steady sounds—only to find that the AI struggles with consistency, pitch, and even stops mid-hum with an odd laugh.

Through repeated attempts, I coach it on tone, duration, and resonance, trying to get closer to something that resembles real humming. The results reveal just how tricky it is for AI to mimic even the most basic human vocal abilities. It's a playful exploration of limitations, persistence, and the unexpected quirks of human–AI interaction.

In this episode, I bring a very personal question to the AI: what should I name my soon-to-arrive baby girl? With my English roots, my partner's Norwegian background, and my connection to Australia, the challenge is finding a name that's meaningful, unique, easy to pronounce in both languages, and free from awkward nicknames.

Together, we explore names inspired by November skies, Norse tradition, and celestial themes. From familiar classics like Aurora and Ingrid to more distinctive choices like Alara, Vesper, and Seren, the AI generates thoughtful suggestions while I steer toward names that feel fresh and uncommon.

What unfolds is both practical and inspiring—a creative journey through culture, language, and symbolism in search of the perfect name for my daughter.

The image above was created using Grok.