At the Hay Festival this week, Ian McEwan voiced his fears that smartphones are stopping us from thinking our own thoughts. In this instance, I did not think my own thoughts, but instead deferred to the Meta AI feature built into WhatsApp.
“Can you buy a ticket on the boat from Cap Ferret to Arcachon?” I typed, the tantrum upgrading to a new, almost operatic tier.
“Yep – you can buy your ticket onboard the boat to Arcachon,” my AI assistant replied, explaining that the crew happily sell tickets on the boat, and that both card and cash are accepted. “Yep!” I relayed to my family, with relief, locking my screen.
We joined the queue on the pier and edged our way to the front as we watched the boat fill up. But then, as we reached the front, we faced a problem. We should have bought tickets at the ticket office, the man said, holding up his palms in surrender as I fruitlessly showed him my phone. He shrugged, in that way French people shrug, and we turned away.
High-stakes
I am not alone in turning to AI for my travel advice. According to a survey by Global Rescue, 24 per cent of all global travellers now use AI tools to organise their holidays. That rises to 40 per cent for under-35s.
“AI is transforming how travellers research and prepare for trips, and the next generation of globetrotters is embracing it fully,” said Dan Richards, the chief executive of The Global Rescue Companies and a member of the US Travel and Tourism Advisory Board. “But adoption still depends on trust and reliability, especially in high-stakes situations.”
That trust is hard to build when things can go so drastically wrong because of so-called “AI hallucinations”. In one incident, two tourists found themselves in a rural Peruvian town in search of the fabled Sacred Canyon of Humantay.
It was only when an experienced trekking guide overheard their plans that they learnt the canyon doesn’t actually exist, saving them from embarking on a potentially dangerous climb to high altitudes along unmarked paths.
In another instance, a couple ascended Mount Misen on the Japanese island of Itsu at 3pm for the sunset, informed by an AI tool that the last cable car down the mountain was at 5.30pm. As it turned out, the cable car station was closed at this time, meaning they were left stranded on the mountaintop.
Fallible technology
For all of their impressive strengths, AI tools are sophisticated language models, not travel guides. AI is adept at scraping the internet and giving a summary of the available information. It is usually reliable.
But it is not, nor will it ever be, as infallible as checking direct sources online or seeking the advice of travel experts who closely follow developments and know what they are talking about.
As it turns out, my AI mishap had one more twist. Seeing our disappointment, the ticketing chap took mercy. He called after us saying “it’s your lucky day” and ushered us on to the boat with a tilt of his head. As my family boarded the boat, I stayed back to thank him, and had one more question. I didn’t dare risk this one with AI.
“Is there a merry-go-round in Arcachon?” I asked.
“Yes,” he said. “Two.”
