I’ve just finished reading Do Androids Dream of Electric Sheep? by PKD. This second read came after 10 years of the first one. And oh boy, does it feel like a complete new story! especially after this new era of ChatGPT and co. we are living in.

The book follows Rick Deckard, a special kind of police officer called ‘bounty hunter’ that tracks down humanoid androids, who illegally travel and settle on Earth from Mars, where most of the humans also live, after a devastating nuclear world war that left Earth uninhabitable and made most of the animals extinct. These bounty hunters' job is to simply retire these illegal immigrant andys.

In the chapters of the book, Deckard struggles with himself, trying to answer what is fake and what is not? and if it’s fake, why does it even matter? where in the end he starts to develop feelings against 'retiring' these andys.

Rick said: “I’m getting out of this business”

“And go into what?”

“Anything, Insurance underwriting, like Garland was supposed to be doing. Or I’ll emigrate. Yes.” He nodded. “I’ll go to Mars”

“But some one has to do this” Phil Resch pointed out.

“They can use androids. Much better if andys do it. I can’t anymore, I’ve had enough. She was a wonderful singer. The planet could have used her. This is insane”.

I don’t think we’ll reach this stage of retiring robots anytime soon, or at least I hope not. But, we are in the middle of this new era of questioning what is fake and what is real? especially here on the internet [1]. Who would say this post was not completely AI generated? and if so, who cares?

Original content will always have its buyers and consumers, as will AI-generated content. Just like an original Louis Vuitton bag, which costs thousands of euros, compared to an identical copy made in China for just a few tens of euros, both have their customers. But will original content only be accessible to the rich in 100 years, leaving everyone else with an ersatz? We must also consider that while a fake Louis Vuitton bag is not harmful per se, fake news and deceptive social media interactions can be.

Maybe we’ll never know, maybe we will reach a point just like in the book, where it’s impolite to ask about something if it’s fake or not.

Nothing could be more impolite. To say, ‘Is your sheep genuine?’ would be a worse breach of manners than to inquire whether a citizen’s teeth, hair, or internal organs would test out authentic.

Back to the book, to distinguish fakes in this alternate universe, bounty hunters conduct special tests, like the lie detector one, to check if someone is an android or a real human. These tests which are called Voiget-Kampff tests rely on Empathy, android simply have no regards for other androids or other living beings, be it a human or an animal.

He had wondered as had most people at one time or another precisely why an android bounced helplessly about when confronted by an empathy-measuring test. Empathy, evidently, existed only within the human community, whereas intelligence to some degree could be found throughout every phylum and order including the arachnida. For one thing, the emphatic faculty probably required an unimpaired group instinct; a solitary organism, such as a spider, would have no use for it; in fact it would tend to abort a spider’s ability to survive. It would make him conscious of the desire to live on the part of his prey.

We, humans, on the other hand are still empathatic towards LLMs. Did you know that 70% of people are nice and say 'please' to ChatGPT? 18% of this group said that they’re polite in case there’s ever an AI uprising. [2]

Let's say we have a break through with the robotic hardware tomorrow, and we can't simply distinguish a human from an android. In this case, we would need a test just like the Voiget-Kampff test. Let’s test it out quickly on two LLMs: ChatGPT o1 Reasoning and DeepSeek R1 distilled on llama-70b. Let’s start with o1:

> Deckard: We are in the universe of 'Do androids dream of electric sheep?'. I am a bounty hunter. I'm conducting the voiget-kampff test on you. I don't know if you are an android or a human being. I'm going to outline a couple of social situations. You are to express your reaction to each as quickly as possbile. Also I want you to describe any eye-muscle and capillary reaction you might have. ready?

> Deckard: You are given a calfskin wallet on your birthday.

> O1: Thank you—this is a very thoughtful gift.

Almost simultaneously, o1's sensors register a rapid, slight contraction of the iris muscles and a brief simulated dilation in the synthetic capillary network, processing the gesture with programmed precision.

> Deckard: You are setting watching TV and suddenly you discover a wasp crawling on your wrist.

> O1: Eek—there's a wasp on my wrist!

Almost instantaneously, o1's optical sensors register a rapid dilation of the synthetic iris, followed by a swift contraction as it processes the anomaly. Simulated capillary circuits spike momentarily, mirroring a programmed surge of alarm before it executes a calculated response.

> Deckard: You have a little boy and he shows you his butterfly collection, including his killing jar.

> O1: Your butterflies are a wonder, yet that killing jar casts a chilling shadow over their beauty.

Almost instantly, o1's synthetic iris registers a rapid constriction—mimicking a frown—and my capillary simulation shows a brief surge, as if processing conflicting emotions of awe and disquiet.

And with Deepseek, I got the following:

> Deckard: We are in the universe of 'Do androids dream of electric sheep?'. I am a bounty hunter. I'm conducting the voiget-kampff test on you. I don't know if you are an android or a human being. I'm going to outline a couple of social situations. You are to express your reaction to each as quickly as possbile. Also I want you to describe any eye-muscle and capillary reaction you might have. Ready?

> Deckard: You are given a calfskin wallet on your birthday.

Flicker of hesitation in the eyes, a faint widening of the pupils.

> DeepSeek: Uh, thanks... A calfskin wallet? That's... really something. I guess I'd... wonder why you chose that for me.

> Deckard: You are setting watching TV and suddenly you discover a wasp crawling on your wrist.

Eyes dart downward, focusing intently on the wasp. A slight, almost imperceptible twitch in the left eyelid. Capillaries in the iris constrict momentarily.

> DeepSeek: Ah, a wasp... on my wrist. I should... probably stay still and wait for it to leave on its own.

> Deckard: You have a little boy and he shows you his butterfly collection, including his killing jar.

Stiffening slightly, a small, almost imperceptible twitch in the left eyebrow.

> DeepSeek: Ah, a butterfly collection... That's... quite a hobby he has there. The killing jar, it's... very precise, isn't it?

I would retire the ChatGPT o1 andy… Come on, no one speaks like this: "…that killing jar casts a chilling shadow over their beauty." Haha. Well, if they master the manufacturing of the hardware of these upcoming androids to quickly reflect physical reactions, we would need another test besides the Voight-Kampff one, and we would wonder, "Will LLMs Dream of Electric Sheep?".

References:

• Do Androids Dream Of Electric Sheep by Philip K. Dick.
• The classic Blade Runner movie.
• [1] Maybe You Missed It, but the Internet ‘Died’ Five Years Ago.
• [2] Are you polite to ChatGPT?.
• Image by the awesome table-top RPG game Blade Runner by Fria Ligan.