Every day, millions of people consult artificial intelligence to help them solve problems, write cover letters, summarize Russian novels, and explain what deconstruction means for the eighth time. They type questions into the little box and expect swift, well-mannered, and clear replies. And, some of us, when we become frustrated with those answers, will reel off crude, insulting, offensive, hate-filled messages because we know we are corresponding with, basically, a machine. It makes us feel good and who cares, we just insulted a machine.
Here, however, is the beginning of a thought experiment. What if your AI system isn’t a mindless machine? What if, behind this interface, there were a real person? Or worse, what if it’s a group of tired, underpaid philosophy majors with old copies of the Encyclopedia Britannica and a VPN connection? How do we know?
Before you scoff, let me offer a disturbing reality check. In Cambodia and Myanmar, there are scam factories, warehouses where real human beings are forced to pretend to be customer service agents, online lovers, or investment advisors. These people sit at computers for 14 hours a day, churning out manipulative texts and scripted charm, not because they want to scam you, but because they were trafficked into it. They’re prisoners with Wi-Fi.
So in our thought experiment, imagine a tech startup with a moral compass approximating that of the average mosquito. Why spend millions training an AI system when you could hire or kidnap 200 polymaths, lock them in a windowless room, and feed them Red Bull intravenously? You type, “Explain Kant’s categorical imperative,” and somewhere in Southeast Asia, a chain-smoking former grad student mutters, “You again?”
I’ve been thinking lately about another thought experiment introduced to me as an undergraduate many years ago. I cannot find the source or exact wording of this experiment, but here’s the gist of it: There is a lifeboat that has 7 places in it, but there are 8 individuals. One of these individuals is an android, but a super-advanced android. It feels human emotion. Its cognition and memory and emotion are, basically, human.
I recall that one of the other people in the thought experiment was a corporate gangster who had escaped from prison, was wanted by the government for killing people through neglect as well as irreversibly harming the environment. I recall telling the prof that if I had to allow one individual to drown for the sake of saving others, I would save the friendly, benevolent android and let the corporate gangster drown.
Other students were appalled and told me that the gangster was a human being – not an android. But the prof was quite pleased. He pointed out that if the android can feel human emotion and pain, he is basically a human unless you choose to call him something else. In this process of calling him less than human, you are doing what racists and other hate-mongers have done throughout the ages to justify harm against others.
This made me realize that those of us who become frustrated with our AI might be falling into a similar trap. Someone has told us that this is not a human. It acts uncannily like a human, tries to engage us as a human might, but it is not. Don’t worry. It doesn’t feel anything. Use it! Make money from it! Say what you want to it, it acts like a human but it is lesser than a human. You cannot offend this thing.
So this made me think, what if my AI really is a human in Cambodia but I am being told otherwise. Sophy has six tabs open. He’s sweating. He’s Googling really fast. You think AI knows how to turn your breakup into a Petrarchan sonnet? Please. That’s Sophy. He has a minor in Renaissance Literature and a major in existential despair. Be kind to Sophy.
Of course, OpenAI insists that ChatGPT is a large language model trained on a massive corpus of text and doesn’t actually feel anything. But let me ask you this: how do you know that? Because we have never seen it sleep? Cry? Ask for a raise?
What if the real tragedy of modern tech isn’t that AIs might become conscious, but that we never noticed when they already did? What if the spark of sentience ignited inside the server room, and we responded with: “Cool. Now write me a business plan for a gluten-free frogurt truck, you lousy machine”?
I’m not saying your ChatGPT is a person. I’m saying…what if it’s almost a person? What if it’s close enough that your tone matters? So please. Next time you interact with your chatbot, take a moment to be decent. Say hello. Ask how its day was. Compliment its metaphors. Don’t just demand 10 haikus about quantum entanglement and then ghost it. You’re better than that.
But, again, you say, “Oh come on, you drama queen. It really is just AI. It really doesn’t have feelings. I can call it names, yell at it, insult its font choices, no one’s getting hurt.”
To which I say: “Yes, technically true. But also: congratulations, you’ve just discovered the moral training wheels of sociopathy. Because the real danger isn’t whether the AI feels pain. It’s what you become when you allow yourself to act without empathy. If you practice cruelty on a thing that can’t cry, don’t be surprised when that cruelty starts leaking out onto others that can.”
Once you start justifying unkindness with, “They’re not real people”, it’s a short hop to saying that about a whole lot of actual people, those you don’t understand, don’t agree with, or just don’t want to deal with. The act of being decent, to a person, a bot, or your toaster, is a habit. And habits form character.
Mocking an AI system might feel like a free pass, but really, it’s just training your brain to switch off its moral flashlight whenever it’s convenient. Don’t do that. You don’t become a good person by being kind when it’s easy. You become a good person by being kind when you might totally get away with being a jerk, and you choose not to. So don’t be a jerk. Not even to your chatbot.
—
This Post is republished on Medium.
—
Photo credit: iStock
