ChatGPT, from OpenAI, is simple: Type in a request, get back a human-like response. Example: “Write a 500-word essay detailing the origins of the Spanish language, its ties to Latin roots, and the pervasiveness today.” Or “Write a warm, humorous 150-word email asking for 30 minutes of the recipient’s time.”
When my students at Stanford Business School demonstrated the product in the classroom last week, I feared for my job. Writing emails and memos is what I teach.
I found the AI-generated work amazing — then breathed a sigh of relief. These early attempts at AI, at best, automate mediocrity. And sometimes they get it wrong entirely.
My former student Matt Gibstein, now founder of an investment firm backing consumer technology start-ups, asked the chatbot to “Write an email to Glenn Kramon, a lecturer at Stanford and newspaper editor, in a tone that is persuasive and likely to make Glenn inclined to take a brief phone call. Please reference a specific article he has written and say how much I enjoyed it.”
I laughed when I read the result. The email began, “I hope this email finds you well.” As my 2,000 former students will tell you, never start an email that way. Beginning with “hope you are well” makes you one of a million; you want to be one in a million.
The email continues: “I am a huge fan of your work” — also a sycophantic cliche — and mentions my “article on the rise of automation and its impact on the workforce,” calling it “incredibly thought provoking and well-written.” I have written no such article (although I did oversee a project on this subject seven years ago).
It continues: “I am reaching out because I am currently working on a project ….” (My students would snicker; unnecessary adverbs — particularly “currently” — are my pet peeve.)
But mainly what automation has failed to master is warmth and individuality, which are essential to persuading most people to answer you. Indeed, now that many software products are helping us write, a personal touch has never been more important to distinguish yourself.
Matt and I receive scores of emails a day and we can easily tell when one is impersonal, and written with a software assist.
This is not to say that ChatGPT (GPT stands for generative pre-trained transformer) isn’t pretty good at some things. It’s clear and grammatical. It’s organized and uses transitions.
Matt suggested we ask the chatbot to write this very article: “Write an 800-word opinion piece about what ChatGPT is, as well as its benefits and drawbacks.” In seconds it delivered a sensible, fair-minded description.
Among the benefits it mentions: “its ability to generate large amounts of text quickly and efficiently. Unlike a human writer, who may struggle to come up with ideas or struggle with writer’s block, the ChatGPT can churn out pages of text in a matter of minutes.” It also praises “its ability to mimic the styles and patterns of different authors.”
Among the drawbacks: “It is not capable of understanding the meaning of the words it generates.” Another challenge is it “lacks the creativity and originality of a human writer …. This can make its output feel repetitive and uninspired.”
No doubt ChatGPT and tools like it will become much more valuable and human-sounding — especially when these products ingest and learn from millions of emails and texts to create more personalized responses. In some future world, AI may be at the head of writers’ rooms, classrooms and law practices. Perhaps someday it may even write complicated code.
For now, if all you are aiming for is “just OK,” these tools will suffice. If you want to stand out and make a personal connection in your emails and essays, you’ll still need to write them yourself. No program can replicate your voice, quirks or wit.
My best advice to those ready to cede to the computer: Don’t outsource your writing to an AI overlord. At least, not yet.
Glenn Kramon is a lecturer on practical writing at the Stanford Graduate School of Business.