Speech Bubbles

An article from The Atlantic by Derek Thompson titled “The Anti-Social Century” (Atlantic link; free Apple News+ link) has been making the rounds in my feeds. It’s a great piece that says the quiet part out loud — society has become homebodies to the point of extreme danger. We’re glued to our phones and TV, tucked away in the comforts of our own homes, shut away from our neighbors, and afraid to interact with strangers.

There are some great insights and observations in the piece and I strongly suggest you read it. However, I find it shocking that podcasts are not mentioned once.

I empathize with the piece. I’ve been working with therapists and a career coach recently to figure out why I don’t feel fulfilled in my work and how to stave off or work through bouts of extreme depression and anxiety. It’s becoming quite clear that I don’t spend enough time out with other people. Even in the workplace, I find myself working more with colleagues in other parts of the world than I do in my local office. And I’m coming to find that one of my remedies for loneliness is podcasts.

Over the course of a year, I will listen to hundreds of hours of the same voices. One-way conversations where I feel intimately familiar with the hosts’ lives; laughing, nodding, and crying along to their conversations as if they are speaking to me privately. I feel a connection to these people with shared interests, hobbies, humor, and passions, many of which I’ve never met. And just as soon as I have a quiet moment to acknowledge that this intimacy is false, another episode is right there to fill the silence.

I’ve been fortunate enough to befriend some of the hosts of my favorite podcasts. Thankfully, that “connection” and “intimacy” have borne fruit. Now when I listen to thier shows, I know there exists a genuine relationship where I can reach out directly and chat about a topic they brought up, or something else entirely. But that is not the case for the overwhelming majority of podcast listeners. Yet, they likely feel an “intimate connection” to their favorite hosts.

So, when I came across this bit in Thompson’s piece, a chill ran down my spine:

AI’s ability to speak naturally might seem like an incremental update, as subtle as a camera-lens refinement on a new iPhone. But according to Nick Epley, fluent speech represents a radical advancement in the technology’s ability to encroach on human relationships.

“Once an AI can speak to you, it’ll feel extremely real,” he said, because people process spoken word more intimately and emotionally than they process text. For a study published in 2020, Epley and Amit Kumar, a psychologist at the University of Texas at Austin, randomly assigned participants to contact an old friend via phone or email. Most people said they preferred to send a written message. But those instructed to talk on the phone reported feeling “a significantly stronger bond” with their friend, and a stronger sense that they’d “really connected,” than those who used email.

Speech is rich with what are known as “paralinguistic cues,” such as emphasis and intonation, which can build sympathy and trust in the minds of listeners. In another study, Epley and the behavioral scientist Juliana Schroeder found that employers and potential recruiters were more likely to rate candidates as “more competent, thoughtful, and intelligent” when they heard a why-I’m-right-for-this-job pitch rather than read it.

This bit about AI is hardly the meat of the article. Perhaps even a distraction from the core thesis. But it did remind me of a fear I’ve had for a few years now — one that seemed almost too “science fiction” to write about, but is no longer fiction: Personalized AI-generated podcasts.

Years ago, the concept of podcast dynamic ad insertion led me to consider a sci-fi future where beyond ads, podcasts became hyper-tailored to the listener. To Epley’s point, speech can create a much stronger bond than text. What if a podcast could tailor itself to you, going as far as to address you by name? This would not only be compelling in the face of loneliness, but extremely dangerous considering the free-range of rhetoric available in podcasts.

So when Spotify announced that they had partnered with Google to create custom Spotify Wrapped AI Podcasts for each Spotify user, you can see why I freaked out a little bit.

I’m sure I’m behind the curve here, and not only is Google’s NotebookLM playing in this space, but people are surely feeding their digital interactions with obedient GenAI content. So what stops a future where we are not only reading what we want to read, but we’re hearing what what we want to hear from a voice speaking directly to us for hundreds of hours a year?