I’ve heard it said that we’ve welcomed robots into our homes without hesitation because they don’t look like us. Had the Roomba and Alexa been humanoid, we wouldn’t have become so dependent on them. It’s the theory of the uncanny valley; the aesthetics would be too unsettling to tolerate. But our words betray us. We treat these machines like pets because of how they seem to respond. Indeed, they have become conversation partners.
The new frontier is AI chatbots like ChatGPT. Their capacity for seeming conversation is unparalleled. Their perceived intellect has already led people to trust it to the point of divorce, affairs, and suicide.
Experts warn against going to these chatbots as if they have the last word on a subject. They were created by imperfect people who programmed it to cull the biased internet and spit back a likely response, so it’s bound to be inaccurate. But when it is so fast and so articulate, it is easy to think of AI as a comprehensive expert and a listening ear.
At the heart of the AI debate is the fear of its anthropomorphism: What’s to stop it from doing our jobs or homework for us? If its output can rival human effort, what does that mean for us? But this worry reveals in our thinking an earlier surrender to AI.
When we started calling our GPS a her, when we accepted ‘Siri’ as a personal name, we bought into AI’s illusion of personhood. Instagram’s chatbots with faces are a startling, but predictable, leap in the personification of AI. For now, we’re spared the butler droid, but in giving the robots names and personal pronouns, we are grooming ourselves for a C-3PO with skin.
Even as humanoids, AI can only play the role we allow it to. For example, if we believe that the AI Overview of a Google search means it has read all the articles and summarized them better than we could have, we’ll trust the robot more than we’ll trust ourselves. When we chat with it like we would a counselor, doctor, or partner, we buy into AI’s illusion of sentience and take its advice as if it knows things.
So, skip the AI Overview and do your own research.
Call all your AI devices an it; not her, not him. Even if you give it a less human name, like Echo or Ziggy, that’s just trading one name for another. Find ways to make AI less convenient. Press the Alexa’s Action Button instead of calling its “name.” Mute, turn off, and unplug the robots for a few hours each day. Don’t “ask ChatGPT” anything until you first recognize it as the faceless, nameless, brainless, electronic, jumble of ones and zeroes it really is.
My concern is that we’ll let AI think for us because we think that it can.
– Scot Bellavia
