Making the most of language generators such as ChatGPT
Like other significant moments in technological development as that development relates to knowledge creation – think Wikipedia in 2001 – the rather sudden and widespread accessibility of artificial intelligence has people freaking out. In higher education, the reaction has been simply to run naked through the streets screaming in primal fear.
Causing this swerve toward frantic freakouts is ChatGPT, which, according to itself, is an AI-driven language model that generates conversational, “humanlike responses to a wide range of inputs, making it useful for applications such as customer service chatbots and social media bots.”
In response to simple queries, ChatGPT can “write” or otherwise cobble together short stories, essays and poetry; evaluate and offer critiques of arguments, speeches and, well, virtually anything that can be dumped into its query box; and it can offer human-sounding answers to questions as far-ranging as “How do I fire someone?” or “Am I a narcissist?”
Imagine Siri and Alexa leaving the house to get their Ph.D.s in everything, then coming home to live in your computer and wait for your next writing or knowledge task. The examples off of the top of my head are merely to highlight just how robust this newish school-slayer is and, with each and every query posed to it, is becoming. That’s right, like HAL 9000 in 2001: A Space Odyssey, ChatGPT is getting ever smarter.
And ChatGPT is not alone. Google is reportedly working on a similar application that promises to be even better, a chatbot called LaMDA, while Microsoft is mulling a $10 billion investment in the technology.
To query or not to query
I first learned about the technology listening to The Daily podcast from the New York Times in which Times tech reporter Kevin Roose walked listeners through a few queries. Host Michael Barbaro asked for a short story about love. Pleased but not overwhelmed, Roose next queried for another short story on love, but this one in Shakespearean iambic pentameter. Finally, Roose asked for the story to be written in the style of a 1940s mobster complaining of the rain. The near-instantaneous results were impressive, though none is likely to win a Booker Prize.
In parallel with augmented reality (AR), AI promises to fundamentally alter how information and knowledge are created and shared. The same sort of intelligence generation that ChatGPT achieves is widely available also in endeavors as disparate as interior design (https://interiorai.com), digital art (https://starryai.com/) and selfies (https://prisma-ai.com/lensa).
You might be skeptical, and that skepticism has likely helped you avoid a lot of misguided and ultimately short-lived technology “revolutions.” You never had a Friendster or MySpace account, you never got duped into paying for TiVo, and somehow you avoided buying a Nook each and every time you entered a Barnes & Noble and had to find a way around that enormous in-store display.
ChatGPT, however, signals that accessible, easy-to-use AI language generators that mimic human conversation aren’t leaving any time soon. Siri and Alexa came back from grad school with luggage.
The applications for customer service, marketing and communications, job training, computer coding and website design, performance reviews and recommendation letters, speech writing, doing market research, and translating from one language to one or more others should be fairly obvious. (Yes, ChatGPT is polylingual, too.)
Just to peak behind the curtain a little bit, I asked ChatGPT, “Why is midcentury modern furniture still so popular?” Here’s the answer I got, in maybe two seconds:
Midcentury modern furniture is still popular because it is characterized by its clean lines, minimalist design, and functionality. These elements are timeless and continue to appeal to people of all ages and styles. Additionally, the style is often seen as a representation of a simpler, more optimistic time in history, which can evoke feelings of nostalgia for some people. Furthermore, the increasing interest in sustainability and minimalism in recent years has led to a renewed appreciation for the simplicity and durability of midcentury modern design.
Pretty good, right? And here’s the thing: If you were to ask ChatGPT the same question, you would get a slightly different answer. The generator isn’t plagiarizing; in other words, it is, in fact, “generating” language. This is why academics are freaking out, because it isn’t cheating as that term historically has been used. It is easy to predict that the bot will be busy this semester cranking out roughly B+ work for students seeking to save time and skirt assignments.
But these are my problems. For furniture retailers, manufacturers, suppliers and reps, ChatGPT promises to take some of the tedium out of routine, mundane, pedestrian tasks and undertakings.
It will be something like hiring an above-average wordsmith and creative, but without demands for pay raises, complaints about coming in on a Friday, or threats to unionize.
For example, imagining that I had been asked to present to my franchisees some ideas for drumming up in-store traffic, I asked the bot, “What are some ideas for creative promotions in home furnishings?”
Here’s half of the decent list I immediately received:
+ Partner with a local interior designer and offer a free design consultation with purchase.
+ Host a furniture makeover event where customers can bring in their old furniture and
receive a discounted rate on new pieces.
+ Offer a “trade-in” program where customers can trade in their old furniture for store
credit towards new purchases.
+ Host a “friends and family” sale where customers can bring in their friends and family to
receive additional discounts on purchases.
+ Collaborate with a home staging company and offer discounted furniture packages for home sellers.
While there is nothing mind-blowing here, if the purpose was simply to kick-start a deeper discussion and brainstorming session, the simple query has to be measured a score. The better and more specific the query is, the better the answer will be, and my queries for this column are necessarily a bit vague. As examples, they are meant to do what I believe the bot will be best used for, which is to get the creative process started.
How about another example to illustrate just a few of the potential applications. I asked the ChatGPT how to insert an image into a webpage using HTML code. The answer is elegant, accurate and – ooh! – illustrated.
One last example, because I simply couldn’t resist the temptation: “ChatGPT, write a headline for this column.”
Now, to be fair, I wrote the actual headline and deckhead BEFORE I queried the bot, so no rotten tomatoes on this one. I’ll let you be the judge: Who (or what) came up with the better headline? Me or the bot?
Here are a few of the 10 suggestions I received:
*How ChatGPT is changing the way businesses operate
*Streamlining customer service with ChatGPT
*The future of AI-powered communication: ChatGPT in the business world
*Boosting productivity and efficiency with ChatGPT
*Revolutionizing the customer experience with ChatGPT
(Please pick the human!)
The lesson for me is that for most of these sorts of queries and activities, the human touch, which is to say, an editor, will continue to be the difference between ChatGPT’s pretty good answers and what we will need to achieve excellence or first-of-class. It is a robot, after all. Like a Roomba, we will still need to come behind and get those corners, do the stairs ourselves, and scoop up the pet hair. (By the way, I don’t have a vacuuming robot, so I asked ChatGPT what these things fail to do well. Ha!) ChatGPT is a tool and most often a starting point. For me, the hardest aspect of a creative task is liftoff, getting started, or moving from inertia into a productive gallop. I can see this tool reliably providing that initial lift.
For my own work responsibilities, I will be deploying ChatGPT to help me write letters of recommendation, appeals of my health care insurance provider’s many denials of coverage, and policy drafts for some of the many committees on which I serve. Not to write them, mind you, but to help me get liftoff. And this is just scratching the surface, so I hope we share the increasingly amazing things we get these AI chatbots to do for us.
Now if you will excuse me, I have to take off all of my clothes and join my colleagues in protecting the ancient root systems of knowledge and retaining walls of academic integrity. Once more unto the breach, dear friends.
Into the breach!