The generative AI phenomenon has spawned a multitude of newsletters, websites and podcasts that promise insights, news, tips, and techniques for understanding and using generative AI as the field moves frenetically forward.
A colleague recommended AI Breakdown so I searched and scrolled through the listings on my podcast app (Overcast – https://overcast.fm – free, easy to use, no ads) to find it (https://agibreakdown.podbean.com) and subscribe.
Over the next three days I listened to a series of short (4-5 minute) chats between Meghan and Ray who each day select a recent scientific publication on AI to discuss. Examples: “Language Models can be Logical Solvers” (http://arxiv.org/pdf/2311.06158v1) and “Engineering a Prompt Engineer” (http://arxiv.org/pdf/2311.05661v1). Quick fire, congenial, conversational, reporting the key findings of each study and the implications. The papers are very up to date – published just a few days the podcast – and extremely technical – impossible for me to understand in the original but moderately comprehensible via the explanations from Meghan and Ray. I was impressed.
On day four I noticed a certain uniformity to the repartee and style. Only then did I consult the show notes:
“Although Megan and Ray lend their voices to the podcast, the content and audio are produced through automated means. Prior to publication, they carefully review the episodes created by AI. They leverage advanced AI technologies, including cutting-edge Large Language Models (LLM) and Text-to-Speech (TTS) systems, to generate captivating episodes.”
Yes, this podcast is AI-generated. And yet, the voices sound natural and the conversation fluid, relaxed and light. The hosts have no doubt cloned the voices of Megan (Maghami) and Ray (Mehran) who are human. Or are they? And have edited the AI-generated text. Did they? Doubts creep in. One more illustration of how far this amazing technology has come.
I reported back to thank my colleague. He pointed out I had chosen the wrong podcast. The AI Breakdown that he follows (https://www.youtube.com/channel/UCKelCK4ZaO6HeEI1KQjqzWA) is on youtube – “daily videos that help you understand the most important news and discussions in AI”. Might be good but I haven’t watched it. And I have not yet checked that the host is human or in silico. Does it even matter?