What ChatGPT Can’t Do

I’d forgotten nearly everything about my years in school, which, given my mild allergic ­reaction to sclerotic and coercive bureaucracies, is merciful. But one thing I do remember: a day, in fifth grade, dedicated to—the future.

Carrying a copy of Jules Verne’s Twenty Thousand Leagues Under the Sea, Tziporah, our no-nonsense homeroom teacher, marched in, hushed us down, and started reading. After twenty wonderful minutes of gushing at everything from Captain Nemo’s lavish quarters to the Nautilus’s advanced machinery, spitting out streams of water as it cut through the ocean like a giant sea monster, Tziporah put down the book and handed us each a sheet of paper and a pencil.

“And now,” she told us, “please draw the future as you imagine it.”

I have no recollection of what I drew—probably some giant robot or a flying car—but one thing stuck and stayed with me for years. However divergent our artistic capabilities or our visions, my classmates and I all had one thing in common: We all imagined a ­future in which some sort of machine does something to make human life a little bit easier and more pleasant.

I thought about that day in fifth grade the other week while playing with ChatGPT, the artificial intelligence chatbot that is ascendant. It has been rewarded with a $29 billion valuation, received its own dedicated parody on South Park, and accumulated a head-scratching 100 million users, a feat that took even mega-popular applications like Instagram years to achieve.

If terms like “chatbot” and “artificial intelligence” are as foreign and as menacing to you as the devilfish were to the Nautilus, a brief explanation is in order. ChatGPT is a piece of software that simulates human conversation. It achieves this miracle because it was fed over eight million documents, was taught ten billion words, and mimics the human brain in its ability to collect ­data, detect patterns, solve problems, learn, and evolve. Or, I should say, mimics it up to a point. Endowed as it is with a central processing unit that is King Kong to a human brain’s Fay Wray, the machine is faster and more efficient than even the brightest of humans. The machine doesn’t get distracted by a video of an adorable otter that pops up on your screen. It doesn’t get up and mosey over to the fridge for a snack. It doesn’t get tired. Anything we can do, it can do faster.

Which, at first blush, is a touch menacing.

One dark night not long ago, I asked ChatGPT to pretend that Groucho Marx was asked to give a Shabbat sermon at his local synagogue. What would the great Groucho say? Within three seconds, the machine regaled me with a perfectly anodyne speech—I’ve definitely heard real, live rabbis give worse—peppered with a few jokes that felt like they were ripped right out of Duck Soup. I asked the machine to repeat the same exercise, but now pretend it was Jerry Seinfeld on the bimah, the rabbi’s elevated platform. This time, two seconds were needed before a short but punchy speech was provided, complete with did-you-ever-notice humorous observations, a Seinfeld signature.

None of these amusements, of course, was meaningful in any way. But imagine asking ChatGPT to scan through, say, 1,400 pages of corporate documents and compose a legal brief. Or digest massive datasets and deliver a scientific paper. Or read every murder mystery ever published in the ­English language and write a novel-length thriller. The machine can already do all that reasonably well, infuriatingly fast, and without human labor. And it’s only getting smarter.

Entertain these and other scenarios, and you’ll see that my fifth-grade fantasy is about to turn into a nightmare. For centuries, human beings imagined the march of progress would lead to paradise. Machines, we daydreamed, would soon do all the manual labor, giving us as much time as we want to think and create. ChatGPT makes it all too clear that the opposite may very well soon be true: Machines will think and create while humans do the manual labor of feeding them data. And with each small improvement in artificial intelligence, our manual labor will be worth less and less.

That, at least, is the opinion of some of our tech overlords, like Elon Musk. By 2025, he opined, artificial intelligence could get so good that machines might—fetch the smelling salts—take over the world. At which point, the Twitter and Tesla boss stated with his characteristic subtlety, things could very soon get “unstable or weird.”

If true, the turbulence may prove to be a minor ­miracle.

Even if you’re willing to sell the human brain short—it’s still a miracle of creation that is infinitely more complex and sophisticated than even the greatest computer at our disposal—you’re still left to reckon with that other endowment that makes our species the glorious animals we are: our souls. Were machines to grow so advanced as to outperform our lawyers, our doctors, and all other highly paid professionals, humans will have no choice but to learn how to be human again, which is to say, how to be soulful and not just clever and smart.

For years now, the art of being human has been lost to many, especially those who take a great deal of pride in being at or near the top of the socioeconomic heap. We hail and reward productivity, focus, and output. We regale ourselves with “life hacks,” ways of being more efficient while exerting less effort. In short, we’ve come to think like machines. Which shouldn’t come as much of a surprise: The reigning metaphors of every age are always closely intertwined with its modes of production. To the soot-stained wretches in the nineteenth century, the factory was the central image, and life was imagined as moving along a production line. Agrarian societies understood themselves in terms of harvests, rainfalls, and droughts. To us, it’s all about rapid computation, performing as many tasks as possible in the shortest time imaginable.

Could there be anything more antithetical to the human soul? Our souls, and our hearts, are trained by sacrifice, for the benefit of others and for the glory of God. The soul is enlarged when we set aside that pile of work emails to play with our small child, when we share our lunch with a hungry person on the street, when we wake up early on a Sunday to go volunteer at the church’s soup kitchen. None of those activities would compute with ChatGPT or its future AI successors. The spirit that moves us will forever remain foreign to machines, however “smart” they become.

And that is wildly promising, not only for our future goodness, but for our future greatness as well. The most foundational human inventions have never been the result of a mere mad dash toward more output. They come, often in bolts of inspiration, from the same stirring spot deep down inside, the place where prayer and hope and love reside. The first human to strike a flint against a rock and make fire; the first human to fashion material into a circle and call it a wheel; the first human to scratch the form of a figure on the wall of a cave so that all fellow humans may bask in its beauty—these are our illustrious forefathers, the pioneers, inventors, and innovators whose genius propelled us from spear to supercomputer in a relatively short period of time.

At each seminal turn, they created not as machines do—by rote, uniformly and unerringly—but like humans. They tried and erred. They thought of their own needs and of the needs of others and took risks to meet them. They found joy in emulating their Maker, creating something where there was nothing before, an act every bit as religious as it is industrious.

For too long, we haven’t seen this magnificent spirit at play and at work. My fifth-grade year was decades ago, and yet I’ve seen no flying cars or lifelike robots, not because we don’t possess the capacity to deliver on such marvelous breakthroughs, but because we’ve focused all of our energy on marching, in a machine-like fashion, toward the maximization of bottom lines. An app that shares photos and makes billions of dollars is a great investment; it’s a calculation that computes. A start-up that delivers a physical, tangible object that actually changes the way we live rather than delivering food that we could easily pick up on our own? Good luck finding someone to finance such a venture.

Maybe now we will see investment flow toward things that engage us spiritually. Maybe we’ll have no choice. ChatGPT and its pals are on course to put many of the information economy’s best and brightest out of business. If and when they do, we’ll find that our finest hour may still be ahead of us, if we allow faith and its soulful ambitions to guide us toward a future brighter than anything we could ever have imagined.

Liel Leibovitz is editor at large for Tablet Magazine and the cohost of its popular podcast, Unorthodox. 

Image by Thomas Leuthardvia Creative Commons.Image cropped.