“Create a travel itinerary for London.” With one simple command, ChatGPT produces five days of suggestions for a trip to London, covering cultural, entertainment and shopping trips.
Day one: historic London, taking in the crown jewels, a walk to Tower Bridge and the Globe Theatre, followed by dinner in the Borough Market area. Day two covers the royal and political scene in London, while the subsequent recommendations are for the more contemporary scene, shopping and relaxation.
Throw in some tips to make travelling easier – weather advice, transport tips and avoiding queues at the more popular attractions – and you have your own personal travel expert at your fingertips.
The only thing it can’t do directly for now is make the bookings for you, though it can tell you the easiest way to do it yourself. But as generative AI is integrated into more websites and services, could that look different in a few years’ time?
Already Google has made changes that will allow its own gen AI, Google Bard, to work alongside Google services such as Google Flights, making it easier to book your summer holiday with a single request to Bard.
If 2023 saw the early stages of generative artificial intelligence, this year is likely to see it develop further. And that could be good news for the planners and organisers among us.
“The benefits hopefully that can be claimed from this kind of technology is that it does free people from more onerous and repetitive tasks,” said John Bowman, AI ethics market strategy lead with IBM. “It allows them to concentrate on work which is of greater quality. But I think where it’s going, if we look at the technological developments like the internet, mobile phones and similar online services, generally it just will become more embedded in everyday life.”
That could mean yet more change for the travel industry. Technology has already caused a significant shift for at least one part of the travel sector – people now book their own trips and compile their own holiday plans rather than relying on a travel agent to do it for them. But there is something reassuring about a travel company taking control and looking after the smaller details of your holiday. Generative AI could help combine the best of both the traditional travel company and more up-to-date technology.
It could help smooth the way for both travellers and the companies involved. Generative AI could mean an end to endless phone calls when flights are cancelled and travel arrangements need to be rebooked.
Imagine a scenario where an AI system could deal with cancelled flights for travellers, not just getting them a new flight but also dealing with any necessary overnight accommodation while they are stranded at the airport. That could be achieved without travellers having to spend hours waiting around an airport or on hold with airline staff, clogging up customer service and reservations lines.
For travellers, it eliminates much of the frustration that comes with this situation. From the airline point of view, it frees up call centres and gives customers a better experience.
At travel company Lonely Planet, AI will open up a whole new facet to its business. The company is best known for its local guides that have steered travellers off the beaten path for decades, with more than 100 titles now in its library.
Generative AI could see that content used in new ways to help travellers.
“Part of what made us want to go down this path is we were looking at our guidebook business. It’s the bedrock of what we do and it’s the product that our customers love. We have 50 years of a corpus of travel content, real recommendations from real people, lived experiences, context,” said Chris Whyde, head of engineering and data science for Lonely Planet.
“It was a big problem to solve: how can we make this content work harder for us? How can we have this show up and experiment with different content formats? Because it’s predominantly a publishing business.”
On the one hand, the company wanted to protect the Lonely Planet guidebook business, which still generates a significant amount of revenue for the business; on the other, it needs to move with the times and get more from the content it has at its disposal.
Striking the right balance between the two is a delicate operation. And so the idea of implementing generative AI was born.
“Everybody was seeing the headlines at the end of last year – ChatGPT got launched and there was a lot of buzz around that. Our engineering team started looking at how maybe this technology could help us solve the problem we haven’t solved yet, which is how can we bring this corpus of content to life in different ways digitally?”
However, like many of the companies currently struggling with fitting the technology into their business model, Lonely Planet wasn’t rushing to hand over its material to generative AI companies to train their technology.
Media companies have begun to take action to protect copyrighted material, with the New York Times among those who are taking action against the AI companies they claim are using content to train the models.
As IBM’s Bowman points out, more and more companies are looking for a way to implement generative AI without losing control of their data, for more reasons than simple concern over intellectual property. In some cases, regulation and data privacy may be fuelling a reluctance to wholeheartedly embrace the new generation of AI tools. But that may leave companies at a disadvantage.
“It’s important for organisations get the balance between being able to innovate, to introduce new products and services or to refine or improve the services they already have, and to manage the risk. So there has to be a risk-based approach to the deployment of AI systems,” he says. “I think that can be grounded within the culture of the organisation. What we found is that it’s important to establish what our ethical principles that underpin the use of AI are, and we can basically anchor our risk-based approach.”
[ Investors to look for AI-powered gains during big-tech earnings seasonOpens in new window ]
For Lonely Planet, which has been working with Amazon Web Service (AWS) for cloud services, the Amazon Bedrock service made the most sense.
“It enabled us to work with our content in a construct that wasn’t training a model; it wasn’t taking that book content and publishing it to the internet,” Whyde says. “So the security and the privacy that AWS was able to bring forward, their approach to generative AI, was a real game-changer in how we were thinking about it.”
Using prompt engineering- plain-English instruction that goes in to the AI model – there was a functioning proof of concept in place within eight weeks that could be tweaked by the Lonely Planet editors.
The system is currently in beta mode but the company plans to have it live for a couple of locations in the coming months. And that, Whyde says, should change things for travellers.
“Planning should be joyous. Travel is life-changing, it sparks joy, and the act of planning a trip should be just as exciting as when you’re about to go on it,” he says. “We just want to take the stress out of it, and we’re really excited with what we’ve seen so far.”
Regardless of how the technology evolves, however, a key consideration for Lonely Planet is maintaining editorial integrity and the general personal experience approach of its guides.
“Consumers will be able to come to a single place, answer a few questions about how you like to travel, and who you’re going to be with, and we will do the rest. It’s the same recommendations stated in the same ways as they are in our books – it’s just a different medium that you can work with,” he says.
“The Lonely Planet product is about real people and lived experiences and that’s not going to change. If that changes, we’re no better than just asking Google what to do.”
- Sign up for push alerts and have the best news, analysis and comment delivered directly to your phone
- Find The Irish Times on WhatsApp and stay up to date
- Our In The News podcast is now published daily – Find the latest episode here
