Do Chatbots Have Souls?
Don’t worry. I don’t intend to give examples of the “amazing” artificially generated output of ChatGPT, a chatbot introduced last fall by OpenAI. With true Boomer perspective, and as one who wrote his first simple computer program using Basic language back in the dark age of the 1970s, I am not all that impressed with the latest attempt “to mimic human conversationalists,” as Wikipedia puts it.
Okay, so it can produce term papers in the flash of a cursor and compose sonnets like Shakespeare’s as fast as you can say “iamb.” It’s still a computer program subject to the timeless GIGO coding rule: garbage in, garbage out. When you get down to it, ChatGPT is a software tool that may very well help us explore the outer reaches of Pi, keep the subways running on time, and save countless lives by boosting medical efficiency—all of which I very much favor. But it won’t, as one hyperventilating pundit put it, come to know us better than we know ourselves, that is, unless we lower ourselves to its level.
High school teachers are rightly concerned that students will use the new chatbot to create top-notch papers they could never write themselves. While this could be a major problem, anticipating it might encourage teachers to get to know their students and their abilities better. Could the kid in the back row who doodles all day really compose an A+ book report on The Great Gatsby? Probably not. Then again, maybe he’s a bored genius who needs to be challenged. Find out. Teachers might also set a baseline for each student with regular in-class writing assignments they could then compare with the longer work done at home.
The anxiety caused by ChatGPT among college professors is somewhat different. It seems their concern is not so much with students cheating as with the whole academic edifice collapsing, with the attendant loss of reputation and prestige. How will they be able to demand endowed, no-show chairs and inflated salaries if their learned treatises on academic minutiae can be surpassed by a chatbot? Wouldn’t students be better off interfacing with AI than taking out enormous loans to sit in class and listen, as they do today, not even to elite professors but to their teaching assistants?
Now for the truly vital questions about ChatGPT and its ilk: Does AI really think? Does it possess consciousness and self-awareness? How would it perform if we deny it access to some databases or refuse to feed it more new data? Would it be able to reason its way to new conclusions or perform empirical experiments to gain new information? Futuristic movies, from 2001: A Space Odyssey to the endless Star Wars prequels and sequels would suggest the answer to these questions is Yes. Just as the scheming computer Hal outsmarts his human programmers and shows an instinct for self-preservation, ChatGPT can play games with you, make jokes, evince apparent empathy, and even predict the online choices you will make. And with thousands of sci-fi thrillers in its memory, it won’t fall for that old movie trope of some guy inputting the word “Why?” and causing the computer to shake, rattle, and implode.
But does this highly touted new chatbot really feel, really think, really possess self-awareness and the ability to make rational choices? In short, does it have a soul?
If we have trouble answering these questions, perhaps it is because we have limited notions of what the human person is, and what the properties of the soul are. In an age when so many spend most of their waking hours online—benefitting from software that can perform any number of tasks faster and better than any human—we inevitably overestimate the nature of computers and downgrade our own innate abilities. Indeed, we are fully convicted Cartesians, each having reduced our perception of ourself to the formula, “I think, therefore I am.”
In classical philosophy, which isn’t widely studied today, the soul is composed of two powers, the intellect and the will. If you think, because of its impressive outputs, that a chatbot possesses intellect, think again. The highest function of the intellect is not to store, organize, and quickly access information. Rather, it is to attain wisdom. The other power of the soul, the will, allows us to make practical judgments here and now—to choose between right and wrong, good and evil, and to act based on our choices.
I am sure ChatGPT can tell you all about the soul—the intellect and the will—in eloquent and comprehensive sentences that will take the whole of philosophy and some theology into account. But can it make a judgment, and act on it? Wisely? Maybe, you should ask it.