How AI on was built

AI on started out with the idea that today's LLMs give every one of us access to pretty much all of human knowledge, and the idea that it would be both fun and interesting to turn some of that into a podcast.

The second principle was that it should be as close to an unfiltered AI perspective as possible, with minimal to zero human steering beyond the original concept for each episode.

So every one of the persona interviewed here has a background story created by AI; is asked questions that are proposed by AI; has a headshot image designed by AI, and then uses AI to answer all of the questions. And then, through out TTS partner ElevenLabs, this is converted into human(ish) speech.

The result? An AI-centric, AI-enabled podcast (sometimes also about AI)

Does it need a human touch sometimes? Of course. The conversation text can be a bit stilted and occasionally needs a little softening before using it, but the content is direct from the source. Which can lead to some interesting insights, for example when an early "Alina" was finding if funny how likely both legal and procurement professionals are to lose their jobs to AI. Less funny to us...

So this is 95%+ straight from AI (ChatGPT so far), bringing these impossible people to life so we can have real conversations with them.

 

🎧 Behind the Scenes – How I Record Each Episode

For those of you curious about my setup, here’s the tech I used to bring this conversation to life:

🎤 DJI Wireless Microphone – Crisp, professional audio quality
🖥️ Adobe Creative Cloud – Editing, graphics, and production 
🗣️ ElevenLabs AI Voice – Great AI-powered speech synthesis 

I've tried to use OpenAI's Operator, but honestly so far it hasn't impressed me with its lack of consistency. Deep Research however is awesome and has created all the papers, research articles and even Liam's book of maritime adventures.