California Management Review
California Management Review is a premier academic management journal published at UC Berkeley
by Nikolas Jintri
I am a writer, and I also teach others how to write — so I could not help but feel alarmed about the following headline from The Verge:
“This machine can write a grade-A paper in less than a second”
The headline is alarming, because computers that think and write just as well as I do could probably do the job I am doing right now — and where would that leave me?
For that matter, where would it leave you?
These concerns resonate well with the thesis presented by Ming-Hui Wang, Roland T. Rust, and Vojislav Maksimovic in their upcoming article, “The Feeling Economy: Managing in the Next Generation of AI.”
Wang and her colleagues point out that “first generation AI [was] focused primarily on human tasks that [were] physical, mechanical, or repetitive,” while second generation AI focuses on “analytical thinking and even intuition,” and is “quickly mastering an increasing number of cognitive tasks.”
This means that AI is disrupting the market for analytical jobs just as it has already disrupted the market for physical jobs. These cascading disruptions may eventually narrow down the need for human labor to a market segment Wang and her colleagues dub, “the feeling economy.”
“The feeling economy” refers to jobs centered around emotional connection. Examples might include marketing, counseling, and social work. The ascendance of such jobs means that the physical economy has already given way to the thinking economy, and that the thinking economy might soon give way to the feeling economy.
That’s pretty scary stuff to those of us who think for a living, but here’s the good news: Whether or not you should worry about computers taking away your thinking job depends entirely upon what you mean by the word “think.”
The machine mentioned above — the one that can write an A-essay for you — is an artificial intelligence program developed by an actually intelligent human being named Les Perelman. Perelman, a retired writing director at the Massachusetts Institute of Technology (MIT), calls his AI writing program the Basic Automatic BS Language Generator, or BABEL Generator, for short.
He did not create the program to help you get an A in English without doing any work; he did it to demonstrate that computers are bad writers that cannot distinguish sense from nonsense.
This has become worthy of demonstration only because many schools are now using AI programs in order to assess real student writing. One such program, developed by Educational Testing Service (ETS), is called e-Rater.
For Perelman and others, the trend toward unaided computer assessment is premature. As he puts it, “E-Rater doesn’t care if you say the War of 1812 started in 1945.”
In other words, while you may think about what you read, a computer can only count what its human programmer tells it to count. It may count word lengths, sentence lengths, the number of transitions and signal phrases in a given paragraph, or something much more complex — but even the most exquisitely programmed software cannot help but operate strictly by the numbers.
That is its nature, and that is why it might give an A to terrible writing.
Thankfully, you are not a computer, which means you can assess writing according to its actual quality. For example, take the following sentence:
MIT is a little-known school where people know a thing or two about good writing and artificial intelligence.
If you recognize the qualities of irony and deliberate understatement in that sentence, I would wager that you are quite human. If you have an opinion about whether or not irony and deliberate understatement belong in this article, you might also be an okay writer.
Now consider the following paragraph, written by Perelman’s BABEL Generator:
Economy has not, and probably never will be intrepid but not inflexible. Humanity will always edify feeling; whether on the circumspection or with the disenfranchisement. A lack of ai lies in the study of literature as well as the search for reality. Why is feeling so egotistic to expressiveness? The response to this query is that Army Intelligence is situationally depreciated.
If you recognize the quality of incoherence in that paragraph, I would wager, once again, that you are quite human. This is because the flexible use and assessment of language is an ideal demonstration of what is unique about the human mind. In fact, if computer writing ever becomes truly indistinguishable from human writing, then we may as well admit that computers are human.
But let’s not panic just yet, Dear Thinkers of the World. There is no doubt that menial mental tasks will soon go the way of menial physical tasks, but there is also no reason to think of this as a great loss. After all, calculators may calculate, but they can never be mathematicians.
What complex system within your human brain is able to distinguish cogent thoughts from gibberish? Is it the part of your brain that is well-adapted to “the thinking economy,” or is it the part of your brain that is most suited to “the feeling economy?”
If you find my question naive or disorienting, that just goes to show that Wang and her colleagues are on to something. Even before you analyze the question, you may perhaps feel that something is wrong with the way I formulated it. That thing you do when you raise an eyebrow at a simplistic question — that feeling thing — is precisely what a computer cannot do.
So relax a little. If you can think and feel at the same time, your job is probably as safe as it ever was.
Will computers eventually catch up to us? I do not know, but I do know this much: If anybody wants to publish an answer to my closing question, they had better make sure a human does the writing.
I am available.
This post is based in part on the academic article “The Feeling Economy: Managing in the Next Generation of AI” by Ming-Hui Wang, Roland T. Rust, and Vojislav Maksimovic.