AI Voice & Blade Runner: What "Human" Really Means
AI Voice & Blade Runner: What "Human" Really Means
A
Andrew D'AmbrosioCo-founder
·
It's funny how some things you learn at school seem completely useless at the time often later become relevant in the strangest ways.
When I was in year 12, believe it or not, the NSW HSC English curriculum offered a study module on Ridley Scott's Blade Runner. Of course, for an 18-year old boy in high school this was awesome.
No-one could have possibly foreseen that we would soon be contemplating the film's core thesis of what it means to be human in real life.
With the latest advancements in AI, science fiction is rapidly becoming fact. In building Voxworks and working with the cutting edge of AI voice conversations, I'm constantly reminded of Blade Runner and have been recently reflecting on some of its main lessons. Let me explain.
More Human Than Human
The Tyrell Corporation headquarters
In Blade Runner, "More human than human" was the motto of the Tyrell Corporation, a company that manufactured artificial humans ("replicants") so sophisticated they were nearly indistinguishable from real people. Somewhat ironically, four decades later, that science fiction tagline has become our industry's aspiration.
At Voxworks, we build AI voice agents that talk with customers over the phone. Every day, we grapple with questions the film raised in 1982: What makes something sound human? Where is the line between artificial and authentic? And what happens when that line blurs?
The Voight-Kampff Test of Our Time
The Voight-Kampff Test was the only way to distinguish between a human and an AI
In Blade Runner, the Voight-Kampff test measured emotional responses to empathetic situations to identify replicants. Subtle physiological cues such as pupil dilation, blush response, micro-expressions revealed the artificial beneath the human facade. Once the replicants revealed their true nature, they were summarily "retired" by the film's protagonist, Rick Deckard, whose job is as an android assassin, a "Blade Runner".
Today's equivalent to the Voight-Kampff test might be called the "phone test." Can an AI carry on a conversation so naturally that the caller doesn't realise they're talking to a machine?
At Voxworks, we analyse thousands of calls to understand where AI falls short. The tells are often subtle:
Response timing that's slightly too consistent
Transitions between topics that feel mechanical
Emotional responses that don't quite match the moment
The absence of verbal filler ("um," "you know," "let me think")
Perfect grammar in situations where humans would stumble
Not quite understanding human nuances like irony or sarcasm
Somewhat perversely, making AI more human sometimes means making it less perfect.
The Uncanny Valley of Voice
Roboticist Masahiro Mori coined the term "uncanny valley" in 1970 to describe the discomfort we feel when something looks almost—but not quite—human. The concept applies equally to voice.
Early text-to-speech systems sat firmly on the robotic side. Nobody mistook them for human, and while they weren't pleasant to listen to, they weren't disturbing either. Modern neural text-to-speech has climbed toward natural human speech, but it occasionally dips into the uncanny valley. In my view, this is why AI voice applications haven't really taken off yet. People have an innate aversion to speaking machines that are almost right but somehow feel wrong. You can't quite put your finger on it.
The tells that trigger this discomfort are fascinating:
Prosodic irregularities: Human speech has natural rhythm variations. We speed up when excited, slow down for emphasis, pause for effect. AI that maintains too-consistent pacing sounds artificial.
Breath simulation: Real humans breathe while speaking. Early AI voices didn't. Modern systems simulate breath, but get it slightly wrong and it becomes unsettling.
Emotional incongruence: When the tone doesn't match the content, listeners sense something is off even if they can't articulate what.
The "uncanny precision": Pronouncing every word perfectly, never stumbling, never self-correcting—this perfection is itself a tell.
Now, when you "cross" the uncanny valley something interesting happens. I've experienced this with some of the most advanced AI voice systems (I've tried almost all by now). You experience a warm naturalness that makes a conversation feel effortless. It's slightly different than talking to a human, because you feel zero bias or judgement, you volunteer things more easily. Moreover, you're subconsciously aware you are directly interfacing with a superintelligence, being the sum total of human knowledge embedded within the LLM.
In my mind, this opens up an exciting array of possibilities for the most common human-to-system interface, otherwise known as customer service. Not that this degree of naturalness is necessarily required to say, book an appointment over the phone, but wouldn't it be nice if it did? And in this new world what other conversations might be initiated with your customers that wouldn't otherwise have happened?
What Tyrell Got Right
The AI executive assistant in Blade Runner, Rachel, was the first artifically-created android that was completely indistinguishable from a human
The genius of Blade Runner was making the viewer question the existential nature of humanity. The main arc of the film revolves around Tyrell's executive assistant, Rachel, who is Tyrell's latest model replicant, the Nexus 6. Tyrell asks: what if a replicant was able to pass the Voight-Kampff test. That is, what if the only way to distinguish replicant from a human no longer works? In one of Harrison Ford's most classic scenes this situation comes to pass, as Rachel passes the Voight-Kampff test and the viewer's mind is spinning as we try to comprehend whether Rachel is really human or not.
Tyrell's replicants were humans by any meaningful definition. They had memories, emotions, relationships, they feared death, they dreamed. If we can recreate ourselves in artificial form, what are we?
By complete contrast, the goal of voice AI shouldn't be perfect imitation of humans. It should be to be to genuine helpfulness in a form people find comfortable.
The best AI voice interactions aren't ones where the caller is fooled. They're ones where the caller's problem gets solved efficiently and pleasantly, regardless of whether they know they're talking to AI.
This reframes the challenge. Instead of asking "How do we trick people into thinking this is human?" we ask "How do we create genuine value while respecting people's intelligence?"
The Ethics of Artificial Voices
Blade Runner was fundamentally a film about our ethical obligations toward artificial beings. Today, we face the inverse question: What are AI's ethical obligations toward humans?
At Voxworks, we've developed principles that guide our work:
Disclosure over deception. We recommend our customers disclose when callers are speaking with AI. Not because we can't make completely convincing AI (spoiler, we can!) but because trust matters more than temporary deception.
Assistance over replacement. Our AI handles routine calls so human teams can focus on complex, high-value interactions. We're not eliminating jobs, we're elevating them.
Empathy over efficiency. Yes, AI can make calls faster. But some calls require patience, understanding, and human connection. Knowing the difference matters.
The Loneliness Problem
One of Blade Runner's most poignant moments is Roy Batty's death monologue. Roy is a being confronting mortality alone, his experiences about to be "lost in time, like tears in rain."
There's an emerging concern about AI companionship: that as AI becomes more conversational, people might prefer AI interactions to human ones. AI doesn't judge, doesn't get tired, doesn't have its own needs. It's available 24/7, endlessly patient, and never has a bad day.
In the business context, we see this manifest differently. Some callers prefer AI because they're embarrassed about their questions. Others because they don't want small talk. Still others because they're calling at 3am when no human is available.
This isn't necessarily problematic. Sometimes AI is genuinely the better option. The risk is when AI becomes a substitute for human connection rather than a complement to it.
The Memory Problem
Replicants in Blade Runner were implanted with false memories, such as photographs and experiences that weren't really theirs. This gave them emotional grounding but also represented a form of manipulation.
AI voice agents face their own memory questions:
How much should AI remember about previous interactions?
Should AI have a consistent "personality" across all calls?
If AI remembers your last conversation, is that helpful or creepy?
Who owns those memories: the business, the caller, or the AI itself?
At Voxworks, we give businesses control over these decisions. Some want their AI to remember returning callers and personalise interactions. Others prefer each call to start fresh. Both approaches have merit depending on context.
Building for the Future
What would Ridley Scott make of today's voice AI? I suspect he'd find it less dramatic than his vision but more philosophically interesting.
Blade Runner's replicants were obviously artificial beings with human characteristics. Today's AI is the inverse, human-sounding voices with obviously artificial intelligence. The gap is closing from both directions.
At Voxworks, we're building for a future where AI handles an increasing proportion of routine communication. We're not concerned about this becoming a dystopic science fiction type situation because we think of AI voice as a more advanced phone answering machine. When AI handles the repetitive, humans can focus on the meaningful.
Calls get answered, questions get resolved, and people get helped. Sometimes that help comes from AI. Sometimes from humans. Often from both working together.
Final Thoughts
"More human than human" was a corporate slogan in a cautionary tale. But perhaps it points toward something worth pursuing: technology that serves human needs, respects human dignity, and enhances human capability.
The AI voice revolution is less about creating artificial beings that pass for human, and more about building tools that help humans be more effective. Businesses that never miss a call, customers who never wait on hold, teams that focus on what matters most.
We're not building replicants at Voxworks and we're not flippant about how this technology potentially impacts people's lives, both positively and negatively. You'll never see us advertising that our AI agents will replace your staff, or pretend as though they are real people on your team. We're respectful of human intelligence and the capacity of each individual to contibute in their own way. At the end of the day, AI is just software and it should serve the interests of people.
At Voxworks, we're building call handling infrastructure for a future where communication works better for everyone. Unlike Roy Batty's experiences or today's antiquated business phone systems, the conversations happening on our platform aren't lost in time. They're transcribed, summarised, analysed and acted upon.
That's the real science fiction made fact. Less human than human, but more available, more consistent, and more scalable than any human team could be alone.
Experience the future of voice AI at voxworks.ai. Start your free trial today.
The Tyrell Corporation headquarters
The Voight-Kampff Test was the only way to distinguish between a human and an AI
The AI executive assistant in Blade Runner, Rachel, was the first artifically-created android that was completely indistinguishable from a human