There’s been a lot of hand-wringing over the socioethical implications of artificial intelligence. Though AI has stoked public curiosity for ages, it wasn’t until the release of ChatGPT, an artificial intelligence language model designed by the San Francisco-based research lab OpenAI, that everyone from statisticians to surgeons started questioning their job security.
GPT-3, the newest and most ambitious version of the chatbot, has demonstrated an uncannily human ability to answer questions in a conversational way. The more you engage with it, the more it personalizes its responses. Leaders in the hospitality and tourism sector have taken notice, as this will inevitably impact trip planning, language translation, and more.
In the most Chicken Little scenario, travel agents, tour guides, hotel concierges, and writers like myself could be out of work in five years (or less). In reality, it’s more likely that AI will help us do our jobs better and more efficiently. To test ChatGPT’s current capabilities when it comes to trip design, I tasked it with planning my next vacation—a familymoon with a newly mobile toddler.
The challenge
After 20 years together, my husband and I finally tied the knot last summer. Our honeymoon has become a family vacation, which means taking the needs and limitations of a 14-month-old into consideration.
Though we already had a shortlist of destinations in mind, I was curious what ChatGPT would suggest without knowing our preferences, budget, or time constraints. After congratulating me on my marriage, which was both sweet and unnerving, it rattled off a greatest hits list of generic familymoon ideas: Hawaii (been there, done that), Disney World (no thanks—we have plenty of time for character meet-and-greets), and the Caribbean (perfectly fine, but it doesn’t feel epic enough for this once-in-a-lifetime trip).
Clearly ChatGPT would benefit from more specificity, so I shared our homebase (Minneapolis), vacation allotment (one to two weeks), max budget ($10,000), and child’s age. It rattled off more bucket list-y picks (Japan, New Zealand, Australia, South Africa), all of which I’ve visited, and seemed to forget that we have a baby in tow; its top reasons for picking New Zealand included bungee jumping and skydiving.
Short of time, I tell it that our frontrunner is the Faroe Islands. ChatGPT spends the next five paragraphs hard selling the remote North Atlantic archipelago as a fabulous destination, playing up the self-governing Danish territory’s “rugged coastlines, picturesque villages, and dramatic waterfalls.” It also warns me that the volcanic islands can be challenging weather-wise, crazy expensive to visit (no kidding), and tricky with a toddler due to its sheer ruggedness. I’m annoyed at ChatGPT’s honesty but also know that it’s right. But these are risks we’re willing to accept.
When to go
When I ask for the best time of year to visit the Faroe Islands, ChatGPT recommends June, July, and August for longer days (up to 19 hours of sunlight) and milder weather. These months, it also notes, are packed with cultural events such as Summer Solstice and the Summarfestivalurin music festival. Unfortunately, the dates it provides for both are off—the former by two days and the latter by two months.
Getting there
Knowing I’m based in Minneapolis (MSP), ChatGPT advises we fly directly to Copenhagen (CPH) and catch a connecting flight to Vágar Airport in the Faroe Islands. Delta, KLM, and Air France operate flights from MSP to CPH, it says, and from CPH I can fly to Vágar Airport with Atlantic Airways, the national airline of the Faroe Islands. The total travel time would be 12 to 15 hours.
This sounds great, except it’s not possible. There are no direct flights from MSP to CPH; most take 35 to 40 hours and require patching through other cities. Flights are outrageously expensive, too, starting at $4,000 for a family of three flying economy and climbing upward of $20,000. More annoyingly, ChatGPT misses the cheapest, fastest routings: six-hour Icelandair or Delta flights direct from Minneapolis to Reykjavik starting around $1,800, followed by a one-hour flight on Atlantic Airways for an additional $825. Airline bookings, it seems, are outside ChatGPT’s wheelhouse.
Getting around
Prices for car rentals depend on when you’re visiting and the type of vehicle you need. The range that ChatGPT quotes me—500 to 1,200 Danish Krone ($70 to $170 USD) per day—is similar to what I find on Expedia and Kayak, but it offers no tips for shaving down the costs. It does, however, remind me that a car seat will cost extra.
Where to stay
We’re traveling with a baby, but this is still our honeymoon and we want to splurge—at least for a night or two. For that, ChatGPT recommends Hotel Føroyar in Tórshavn for its “spacious and elegantly designed rooms with panoramic views of the city and surrounding mountains and sea.” It also suggests the four-star Havgrím Seaside Hotel 1948 in Tórshavn and the charming Gjáargardur Guesthouse in Gjógv, both of which look promising. (Less enticing is Guesthouse Marknagil in Leynar, which has middling reviews). Unfortunately, ChatGPT’s generic property descriptions and a lack of pricing means we’ll have to do our own research. In addition to misspelling the name of the four-star Hotel Brandan in Tórshavn, for example, ChatGPT fails to mention that it’s now Green Key-certified, making it one of the islands’ most sustainable booking options.
When I inquired about Airbnbs, ChatGPT reeled off six ideas. They sounded lovely, but it included no direct links. What listings I was able to unearth based on the chatbot’s hurried descriptions looked grim. If we decide vacation rentals are the way to go, it’s back to square one on the research front. (Sigh.)
What to eat, see, and do
“What are the best things to do with a baby in the Faroe Islands?,” I ask. ChatGPT tells me to take a scenic drive (okay, where?), hop on a boat tour (which one?), visit a picturesque village (um, any suggestions?), and go for a hike (like… off a cliff?). ChatGPT’s lack of specificity is exhausting.
It fared slightly better on the restaurant front, flagging Áarstova, a cozy eatery in the old town of Tórshavn for traditional Faroese dishes, and the turf-roofed Barbara Fish House, also in the capital, for lobster and mussels. Its top pick, however, was the Michelin-starred restaurant KOKS, which cannot accommodate young children and has temporarily relocated to Ilimanaq Lodge in Greenland. (Greenland!)
What to bring home
What souvenirs does ChatGPT think are worth schlepping home? Hand-knit woolen sweaters, hats, and blankets; Faroese delicacies like dried fish, lamb, and honey; and locally made pottery and glassware. When I asked where to buy these things, ChatGPT name-dropped Guðrun & Guðrun, a high-end knitwear boutique in Tórshavn that every publication on earth has covered. It also confused a restaurant for a clothing store, suggested two shop names for which I could find no trace online, and mixed up heimablídni, the delightful practice of Faroese opening their homes to travelers to share traditional foods, crafts, and storytelling, with a grocery store that sells mutton.
Cultural etiquette
The chatbot’s advice: Don’t be late, don’t litter, and don’t get hammered in public. Pretty universal stuff. The one tip that took me by surprise: It’s customary to remove your shoes when entering a home in the Faroe Islands. One up for ChatGPT’s team—this I did not know.
Insider tips
When I ask ChatGPT for the inside track on the Faroe Islands, it tells me to dress in layers for the unpredictable weather (no revelations there); sample the local cuisine (you don’t say?!); and take a guided tour to “access hidden gems and off-the-beaten-path destinations.” Cliché-packed writing aside, there is no direction on which guided tours are the best and what you might learn from them. Stating the obvious, it seems, is ChatGPT’s only party trick.
After conversing with the chatbot for two hours, I’m nowhere near finalizing our vacation plans. It has given me some useful information—namely the best time of year to travel and a few hotels and restaurants to consider—but nothing it recommended can be taken as gospel. Will its ability to provide helpful trip-planning services improve exponentially in the years to come? No doubt. But for now, nothing beats good old-fashioned, human-powered research.