TTE 2020: AI will automate not terminate, says Travelport data analytics head

“Give a benefit to the people you are getting data from then they will share it with you.”

Artificial Intelligence and machine learning is not futuristic Terminator type technology that will threaten mankind, delegates at Travelport’s digital breakfast during TTE were told.

Stephen Oman, director of data analytics at Travelport Digital, said AI “has been around for the best part of 50 years”.

“The problem is it’s been stuck in academia. There have been two problems. The first is there was no data. Today, as we know, Google tracks everything you do online, no matter where you go.

“Most of this is benign, it’s designed so people can figure out what users are doing so they can run their products better.”

“The second change is now there is fancy hardware designed to do all the maths behind AI to make it much faster.

“There are obviously legal and ethical implications about how this data is going to be used. Once you are clear and transparent about who you are then you are going to be okay.

“Give a benefit to the people you are getting data from then they will share it with you.”

Oman added AI is an umbrella terms for different types of technology and he said in order to use it in business “you do not need to understand at a detailed level how it works”.

“It’s just mathematical software. There are lots of technical terms like neural networks and deep learning but that’s just what’s applied to a particular algorithm.

“It’s not going to run away and take over the world.”

Oman set out the five classifications of AI based on how the technology is being applied by businesses and what it is intended to achieve.

The first was ranking, and how lists of items are displayed contextually to the user. The second is recommendations in which AI tries to work out what will interest someone.

The third use case is classification, where the technology figures out whether something is like something else or not, then there is regression, or predicting things like prices in the future.

The fifth use case is clustering, where the algorithm gives structure to large audience data sets, and anomaly, or looking for anything that looks odd, or out of place, like unusual bank transactions.

Oman said AI is capable of doing one or a combination of these calculations much more quickly than any human would be able to.

“There are people who offer manual filtering, but it takes more time and effort. I want the top result to be the most relevant, not the one the company wants to sell me.

Amazon masters AI

“Amazon are past masters at generating data. Everything is recorded with the explicit purpose of getting you to buy more stuff.

“The key is to keep you on the website. The longer you are there the more chance there is that you will buy something.

“Not only that but they do A-B testing of slightly different variations of their algorithms to find out which one converts the most.”

An obvious application in travel, Oman added, are itinerary planning that require relevant recommendations for what travellers can do in a destination and how to get there.

“You will see AI building itineraries automatically. It can do that by looking back at all the itineraries of people going to that place, where they go, where they like to stay, where they rebook.

“All that data comes into play when you start predicting things in the future.”

Voice search and chatbot improvements

Chatbots will become increasingly sophisticated by using classification algorithms to power natural language processing and make interactions less like basic Q&As.

“Language is messy, it’s ambiguous. AI has to figure it all out. As a trend, this is getting better. Natural language processing is evolving toward full conversational based AI.”

Oman cited Amtrax’s virtual assistant Julie as a prime example. “It does actually work. No one has to use the phone at all and this saves me time as a traveller.”

Travelport Digital has also developed a voice search interface for the award-winning easyJet app.

AI will also become increasingly common for disruption prediction so travellers will be automatically advised to arrive early at an airport if security lines are long.

And Oman said regression AI is it already being used by the likes of airfare shopping brand Hoppa which advices users on the when to buy to get the best price.

Autonomous vehicle will also be made possible by applying multiple AI processes and techniques, said Oman.

“It’s becoming more and more applicable across a broad swathe of industries. Anything that can be automated will be.

“This is the history of technology. Anything that’s rote, that you do over and over again you will automate that.

“With AI we have a whole bunch of new techniques that will help us to automate our businesses.”