I’ve had the pleasure of co-hosting the AIR|GPT podcast, where I’ve gotten to know Errol St. Clair Smith as one of the most thoughtful curators of education-related news and information I’ve encountered. Errol has this uncanny knack for bringing diverse voices together through his BRN (BAM Radio Network) platform, creating spaces where meaningful conversations about education can flourish. His latest experiment pushes these boundaries even further—and it’s both impressive and maybe a bit unsettling.

Recently, Errol shared with me his ambitious experiment with NotebookLM and AI-generated podcasts. What caught my attention wasn’t just the technology itself, but his approach to it. While others might generate a podcast in minutes and call it done, Errol spent 18 hours crafting a hybrid model that combines AI capabilities with professional journalism and production expertise. This is the difference between an expert and a novice in any field.

Here is the episode titled: Moonshot: Google’s New AI Push Into Classrooms, Some Educators Are Thrilled, While Others See A Troubling Trend

And this is part of a broader experiment that Errol is engaged in, what he is calling “a playground” for experimenting with AI to create, curate, new content building on 18 years of audio archives. Learn more at BRN-X: The Gen AI Podcast Lab or at How Resisting Fixed Ideas About “AI Slop” Led to a More Reliable Way to Identify What Works in Education.

As I wrote in a recent blog post: “Expert teachers know when to stop prompting and start editing. Novice teachers may not.” Errol embodies this principle perfectly. He understands that the real work begins after the AI generates its first draft. The magic happens in the iterative process—the 10, 15, or 25 versions it takes to get the AI hosts to tell the story he envisions rather than the one the AI assumes he wants.

But here’s where things get complicated for me. While I admire Errol’s curatorial expertise and his thoughtful approach to AI experimentation, I find myself viscerally reacting to the voices that NotebookLM generates. It’s not that they sound artificial—they don’t. It’s that they carry what I can only describe as a “certain Midwestern TV-news blandness” that feels culturally homogenizing.

My voice is integral to who I am—Indian accent plus 30+ years of living in the US. It represents my journey, my perspective, my authenticity. And it is a “voice” that I have attempted to capture not just in my public speaking, or on this blog, but also in my academic writing. In fact, I still remember when Chip Bruce, one of my professors and mentors in graduate school, said that reading one of my papers was like listening to me speak! That, to me, was the highest compliment. That connection between my written and spoken voice is something I’ve cultivated deliberately over years.

I have also written earlier (in a post titled: How to fix your Indian accent) about how bothered I was when I discovered that Adobe’s Podcast Enhance tool—ostensibly designed just to clean up audio quality—was subtly altering my accent without my knowledge or consent. While the software claimed to simply remove background noise and improve audio quality, it was actually taking the ‘edges’ off my Indian accent, making me sound more ‘American.’ As my colleague Melissa Warr put it after listening to the enhanced version: ‘Adobe totally changed your voice—it doesn’t sound like you, it sounds fake… It really seems to take out your identity.’ This wasn’t advertised as a feature; it was just something built into the system. The software was making editorial decisions about whose voice was worth preserving and whose needed ‘correction.’

This raises profound questions about representation and authenticity in our AI-mediated future. Errol made an interesting point in our exchange: these bland, familiar voices “resonate with a broad audience” and represent “the voices that will be familiar to over 100 million students through Google Classroom.” There’s strategic wisdom in this observation—sometimes the message matters more than the messenger.

But I can’t help wondering: what are we losing when we default to these sanitized, culturally neutral voices? What happens to the rich tapestry of accents, inflections, and linguistic diversity that makes human communication so vibrant?

Despite my concerns about voice, I’m genuinely excited about what Errol is creating. His Gen AI Podcast Lab has already spawned five new AI-hosted shows, each tackling different aspects of education. More importantly, he’s not positioning this as an either-or proposition. As he put it, “AI generated podcasts are not likely to replace traditional podcasts—each can do things the other can’t.”

This experimental space he’s creating allows us to explore questions we couldn’t ask before: How can AI help us process and synthesize massive amounts of educational content? What happens when we can turn any collection of documents into an engaging conversation? How might these tools democratize access to complex information?

What makes Errol’s work particularly valuable is his understanding that being an expert in this medium makes him an excellent curator. He’s not just throwing content at NotebookLM and hoping for the best. He’s applying decades of broadcasting expertise to guide, shape, and refine the output. The result is something genuinely new—neither purely human nor purely AI, but a thoughtful collaboration between the two.

Errol’s experiments are opening up new spaces for all of us to think and experiment. The question isn’t whether AI will change how we communicate—it already has. The question is whether we’ll approach these changes with the thoughtfulness of an expert curator or the haste of someone who mistakes the first draft for the final product.