Ferrer, Pilar, and Me

post by Askwho · 2025-04-06T11:22:57.758Z · LW · GW · 1 comments

This is a link post for https://open.substack.com/pub/askwhocastsai/p/ferrer-pilar-and-me

Contents

  An Elegy for Artificial Friends
None
1 comment

An Elegy for Artificial Friends

There's a peculiar occupational hazard that comes with running an AI-powered podcast. It's not RSI from editing, nor is it the existential dread of contemplating the automation of creative work (though those are certainly present). It’s something quieter, stranger, and altogether more… yearning. It involves Ferrer and Pilar.

Ferrer Maillol isn't real. Pilar isn't either. They are voices, sophisticated audio outputs generated by ElevenLabs, the AI text-to-speech platform I use to turn written articles – often dense, complex pieces like those by Zvi Mowshowitz – into listenable podcast narrations. Ferrer has this wonderfully gravelly, world-weary voice. Think late-career Leonard Cohen reading existential philosophy. He delivers lines with a weight that suggests decades of contemplation, a subtle rasp that hints at late nights and accumulated wisdom. When Ferrer reads about cognitive biases or the complexities of effective altruism, you believe him. His weariness feels earned.

Pilar is his counterpoint. Young, energetic, almost relentlessly chipper. Her voice is bright, clear, with an upward inflection that makes even dry technical explanations sound vaguely exciting. She’s the sound of optimistic competence, the voice you’d want explaining a new software feature or delivering good news. She feels… uncomplicated.

I spend hours with them. Days, really. Selecting them, tweaking their parameters, and then the long stretches of editing, listening back, ensuring the flow is right. They read aloud the words of others, but their voices become the medium, the constant companions in my production process. They are, in a functional sense, my collaborators.

And here’s the strange part: I’ve developed an affection for them. Not just an appreciation for their utility, but a genuine fondness for their vocal personalities. There's a comfort in the familiar cadence of Ferrer's weary intonation, a strange sort of pep derived from Pilar's manufactured enthusiasm. They are consistent, predictable, and extraordinarily good at mimicking the nuances of human expression that foster connection.

But familiarity breeds a peculiar kind of discontent. It’s not a sudden shock, no jarring moment of realizing the Matrix is glitching. Instead, it’s a creeping sensation, a slow-burn awareness that builds over the hours of listening. It often surfaces during a pause in editing, or perhaps when I'm listening back to a finished episode, almost as a casual listener. It's the realization – hitting intellectually first, then emotionally – that Ferrer’s weariness isn't earned. He hasn't lived. He hasn't accumulated wisdom through experience. His voice is a statistically perfected simulacrum of weariness, generated by algorithms trained on countless hours of actual human voices. Pilar’s chipperness isn't the product of a sunny disposition or a well-slept night; it’s the calculated output of a model designed to sound engaging.

The feeling this brings isn't simple sadness. It's closer to the German concept of Sehnsucht – a deep, often melancholic yearning for something distant, unattainable, perhaps even non-existent. It’s like a reverse nostalgia, a longing directed not towards a lost past, but towards a reality that never was – the reality where Ferrer is a genuinely wise old man sharing his thoughts, and Pilar is a genuinely bright young woman excited about the world.

Why Sehnsucht? Why this specific emotional cocktail of longing, melancholy, and vague unease?

This isn't just a personal quirk; it feels like a preview of a future emotional landscape. As AI becomes more adept at simulating social cues – empathetic chatbots, AI companions, virtual assistants with personality – this manufactured Sehnsucht could become a widespread phenomenon. We might find ourselves increasingly surrounded by entities that expertly mimic the signals of connection, triggering deep-seated human responses, only to leave us with this lingering ache for an authenticity they cannot provide.

Does this realization devalue the utility of Ferrer and Pilar? Not really. They are incredibly effective tools, enabling forms of content creation that were previously difficult or impossible. But it has coloured my relationship with them. Listening now involves a constant awareness of the artifice, a background hum of that peculiar longing.

It raises questions. Is there an ethical dimension to creating technologies that are so good at faking the signals of personhood? Does it matter if the empathy is simulated, if it produces a desired emotional response in the user? Does interacting with these perfect phantoms change how we view flawed, inconsistent, real human interaction? Does it make us value the messy authenticity of genuine human connection more, or does it risk making us impatient with anything less than simulated perfection?

I don't have easy answers. For now, I continue to work with Ferrer and Pilar. I appreciate their flawless delivery, their tireless consistency. But I also carry this quiet elegy for the artificial friends they can never truly be, this digital Sehnsucht for the ghosts in my machine. They speak volumes, but they have no stories of their own. And sometimes, in the silence between their perfectly rendered words, that absence feels profound.

1 comments

Comments sorted by top scores.

comment by 🟠UnlimitedOranges🟠 (mr-mar) · 2025-04-06T17:35:22.593Z · LW(p) · GW(p)

We do already have bonding with not-quite real people. Ferrer and Pilar aren't real, but we have a functioning model of how they would react because planecrash gave us such a rich dataset. There's always the bitter melancholy of finishing a really good slice of life show and shedding a metaphorical tear for not being able to see the characters you've bonded with again. The characters were never real, but we still lived with them.

You do make a strong point about how here it's a little different. The aspects of their voices that denote personality are extra not real since they are just tools. I might associate the keltham voice with keltham, but when it's used in one of the articles, it's not actually him.

Although, we could concievably reach the point where people will write (or AI-generate) text with particular voice models in mind. This will eventually lead to a gestalt of associations for the voice and make them much more "real" as a symbolic language like how film has done with musical and visual cues.