In late November it will be one year since OpenAI launched ChatGPT.
In the months since, figuring out how to use the large
language model technology behind it and other generative artificial intelligence tools has become a
top priority – in some cases the priority – for companies of all types and
sizes across every vertical.
In travel, the topic
gained momentum in January and then in rapid succession several companies,
including Trip.com,
Expedia,
Kayak
and many others began rolling
out tools built on generative AI.
Now, many months in and as some of the hype has subsided, the
conversation is being framed around risks as much as opportunities.
Subscribe to our newsletter below
In a session at WiT Singapore moderated by Airbnb’s general
counsel for APAC, Darrell Chan, executives from Expedia Group, Wego, Travelport
and Google shared what concerns them about generative AI – and also what
excites them about its potential.
“I’m worried about privacy,” said Tom Kershaw, chief product
and technology officer at Travelport.
“The type of data we deal with in our industry is extremely
sensitive. Amazon, they know my shoe size - whoop-de-doo - what are they going
to do with that? We know passport numbers and all kinds of very important,
sensitive information. And making sure that information is protected but still
allows for seamless check-in and eligibility and those types of things is a
very difficult balance.”
Kershaw also said that data silos – a traditional mechanism to
protect sensitive information in travel – is “the enemy of AI” – so the
industry must find a way to protect data while still allowing it to be
accessible across systems.
One way this technology may be used in travel is as an “intelligent
agent,” capable of making decisions and bookings on behalf of a user – driven by
deep understanding of that user’s preferences and needs. But, asked Wego
co-founder Ross Veitch, what happens if that agent makes a mistake?
“Who is liable when it screws up? Is it the customer for not
paying close enough attention, the person who developed the agent in the first
place?” Veitch said.
We know passport numbers and all kinds of very important, sensitive information. And making sure that information is protected but still allows for seamless check-in and eligibility ... is a very difficult balance.
Tom Kershaw - Travelport
Accuracy – and the risk of hallucinations – are also key concerns
about responses generated by large language models. As an example, Veitch said
that when using generative AI to write descriptions of hotels, his company has
found that for properties that do not have a lot of publicly-accessible
information, the system will “just make stuff up.”
Another concern, one that is potentially more dangerous
than hallucinations, is bias in large language models.
Expedia Group senior vice president of product and
technology Karen Bolda said her company has a governance committee that is
tasked with evaluating its use of AI to ensure it aligns with company values.
“One is fairness - preventing unjust biases - and then there
is, of course, privacy, transparency,” she said.
“Do we understand what this model is doing and why it’s
doing it? And are we being transparent with both our travelers and our B2B
partners about the impact of that?”
In addition to policies and guidelines that travel companies may
develop internally, countries around the world are designing and implementing
AI governance legislation.
Google’s managing director of global business development and
partnerships for travel, Gianni Marostica, said regulation is both warranted and
required.
“I equate it to sustainability – if we get this right, the
world will be a much better place,” he said.
“So it’s in all of our interests to do a better job … so
that collaboration within and outside the industry and also collaborating with governments
and regulators is critical.”
The positive potential of generative AI in travel
Along with talk of risks, panelists also expressed optimism
for the positive impact AI can have in the travel industry, both for travelers
and for companies.
“My primary interest in AI is in revolutionizing the back-end systems,” Kershaw said.
“We all talk about shop – find the cheapest price. In the
end, it’s about servicing. … How can you make changes to air schedules automatic? How can you automatically rebook? It’s the back-end systems that matter. It's making
searching efficient and using all the data we have effectively. You can have a
slick front end, but if what it’s talking to doesn’t understand what’s going on,
you’re … not going to get far.”
Veitch said Wego has been working with AI for some time, but
the release of ChatGPT was a “real inflection point,” and now his company is
building out an “intelligent agent” assistant tool that will be integrated in
its mobile app.
“The idea is to augment the rich interface we’ve already
built, to add to it — so voice, natural language capabilities. The ambition is to
be able to interact with it in the same way you would with a really seasoned
travel agent, with the richness of the Wego UI we’ve already built,” he said.
Panelists also discussed opportunities for generative AI to add
efficiency to business travel, to improve forecasting and other internal
operations at travel companies and to accelerate the capabilities of new
entrants into an industry that has traditionally been very closed off.
“Try it out. There are so many tools out there," Kershaw said. "There used
to be big barriers to entry to be able to do things in travel.
“Now you can spin up systems and start to experiment with
things with very little friction. So I encourage everyone to do that and start
to change the way we think about the travel industry.”