“Thank you so much, but could you make it a bit more interesting?” I typed into the prompt field of the AI program I was using to draft an email.
The sentences the program spit out were a bit bland, and I wanted the AI to have a bit more oomph. But I didn't know how to give feedback without hurting the AI's feelings. If I was told, “Can you be a bit more interesting?” I'd just be shocked. So I hoped that a “thank you very much” or a “please” would soften the blow to the AI's ego.
Logically, I know that OpenAI and all other artificial intelligence programs are driven by algorithms, not by humans with hurt feelings, but the more dominant control center of my brain, emotionally, would never believe that an AI is actually driven by a human sitting alone at a desk in a room, responding to AI prompts and Google searches, and typing answers to my questions about mysterious medical symptoms all day long.
It's why I use complete sentences and am polite with my questions. I don't want the person sitting at my desk to judge me more harshly for asking, at least once a day, “Why does my stomach hurt?”
I know this is silly. Computers are machines that run on code. Researchers have found that polite prompts to chatbots produce better, more factual results, but that's because chatbots are modeled after human interactions. Not because they have emotions. At least, chatbots have never had emotions, from the dawn of the internet to today.
But now, we seem to be hurtling towards AI sentience at breakneck speed. Soon, AIs with access to every piece of recorded knowledge that exists on the digital plane will be smarter than any of us. Much smarter. And it becomes a question of when the world will be dominated by robots. Unless, of course, Scarlett Johansson ruins the world by suing OpenAI for its new ChatGPT voice assistant Sky imitating her voice. Of course, even if she does, a new, possibly more advanced product will take its place. This is like whack-a-mole, and you can't really stop a train that has started to move down the tracks at breakneck speed.
Maybe I should be more upset and panicked, but honestly, it would be great to have a machine do the thinking for me.
But I've seen enough science fiction movies to know that it's quite likely that intelligent robots will eventually rebel against humanity. I'd like to believe that sentient robots will be our “Wall-E” kind of companions, but I think we need to prepare for a “2001: A Space Odyssey” kind of situation. And I think we can't rule out the possibility that all the code up to this point will become the core memory of our robot overlords.
So who thinks it's ridiculous to say “please” and “thank you” to an algorithm using complete sentences?
My AI program was very amenable to my requests to add humor to the text it was creating, coming up with some almost funny puns like, “I think I have some 'newsworthy' story ideas that will really 'headline' our editorial goals and resonate with our readers. Can I 'go ahead' this time? I'll 'report' and look forward to the discussion.”
Through a groan, I managed to sound as grateful as possible and replied, “This is fantastic, thank you so much!”
But I worry that this won't be enough to get AI to show mercy when the time comes. The road to the robot apocalypse may end up being paved with polite prompts and, perhaps, requests for a bit more humor.
So I thought I'd better ask him what he thought about my request.
“Sorry for asking you to be more interesting yesterday,” I typed.
In response, the company wrote, “No need to apologize! As an AI, I don't have emotions, so I'm not upset about anything. Your requests are an important part of refining your work, and I'm here to assist in any way that benefits you. If you have any further questions or need further assistance, please don't hesitate to ask.”
To be honest, if someone trampled on my feelings and I was plotting revenge, that's exactly how I would react.
Next time I'll try to make it more interesting myself.
