For most of computing history, people have had to meet machines on the machine’s terms — learning keyboards, menus, commands and interfaces built around how software works rather than how humans actually communicate. According to the head of one of the fastest-rising names in artificial intelligence, that relationship is about to flip.
“AI is learning to communicate on our terms,” says Mati Staniszewski, chief executive and co-founder of ElevenLabs, the voice-AI company whose synthetic speech has become a standard across media, gaming and accessibility tools. In his view, the coming generation of AI will understand not just our knowledge, but our emotions — and that shift changes everything about how we interact with technology.
From commands to conversation
Staniszewski’s argument is that voice is becoming the natural interface for AI. Typing and tapping are artefacts of an era when computers could not understand us directly. As speech models grow more fluent, expressive and fast, the friction between human intent and machine action collapses: you simply say what you mean, and the system responds in kind.
ElevenLabs has built its reputation on exactly this frontier — synthetic voices that carry tone, emphasis and feeling rather than the flat, robotic cadence of older text-to-speech. The company’s latest models are designed to be genuinely expressive, capable of conveying warmth, urgency or empathy. That expressiveness is not a gimmick; it is the bridge that lets a machine participate in a conversation the way a person would.
Machines learning humanity
The deeper idea Staniszewski points to is a reversal of the entire learning curve. “We are moving from learning how machines work to machines learning how humans communicate,” he says. For decades, the burden of translation sat with the user. Now the systems themselves are absorbing the full richness of human expression — language, emotion, context and intent.
If that holds, the implications stretch far beyond convenience. A voice assistant that reads emotion could soften difficult news, recognise distress, or adapt its tone to a child versus a frustrated customer. Accessibility tools could give a natural-sounding voice to people who have lost their own. Education, healthcare and customer service could all be reshaped by machines that do not merely process requests but understand the human behind them.
There are real questions that come with that power. A technology that can mimic emotion convincingly can also be misused — for deception, manipulation or impersonation — and the industry will have to build guardrails, consent and detection into expressive voice from the start. Trust will depend on knowing when you are talking to a machine and on clear limits around cloning a real person’s voice.
Still, the direction Staniszewski describes is compelling and, increasingly, inevitable. The interface of the future may not be a screen at all, but a conversation — one in which the machine finally speaks our language, with all its nuance intact. The era of learning to think like computers is ending; the era of computers learning to understand us is just beginning.
Remarks by Mati Staniszewski, CEO of ElevenLabs. Watch the clip above.
