Smarts Just Went Up a Notch

Smarts Just Went Up a Notch

There's been an important step change in Artificial Intelligence (AI) and the travel industry is already embracing it.

Everyone remembers that moment when a computer beat a human for the first time at chess. Easy Peasy! Chess, it turns out, is not the benchmark. Apparently, the ancient game of GO is way more complex.

In 2015, Google's AI subsidiary company, Deepmind, created an AI machine called AlphaGo which they taught to play Go. It was given a dataset of 100,000 games to analyse and eventually became competent enough to take on and beat the 18-time world champion, Lee Se-dol.

Last July, Deepmind's new machine AlphaGoZero, beat its predessessor by 100 games to nil, and it did it from scratch.

All it was given were the rules to Go. Then it started playing itself. After three days and millions of games it was able to thrash its dad.

Olaf Slater, Director International Strategy & Innovation at SABRE Hospitality Group, used this example last week at the ITB travel trade show in Berlin to illustrate how AI has suddenly got smarter.

"No longer do AI machines have to be trained over years with tons of data" he said. "Now, with very little data, they can understand what they are looking at and extract relevance from that and then either make predictions or decisions to influence the outcome of something."

One of the things that has fuelled the growth in AI is the abundance of computation power. We no longer need to build mighty number-crunching supercomputers. Instead we can go to the cloud where processing power is as cheap as chips.

This puts AI in the hands of ordinary users and small businesses to use as a tool, and in the travel industry it's already happening.

"If you use some of the large OTAs (Online Travel Agencies) - Expedia for example - you might not know it, but you are seeing very good examples of machine learning AI in practise, because it is looking at your behaviour," says Slater. "It's saying Aha, you've been looking at this destination for 3.5 minutes, and these four hotels, and when you looked at the first hotel you looked at this part of the description... I think that what you are looking for is actually this, and it will change the display accordingly."

"Hooray! Result!" says John Musk, Product Director at TravelRepublic, who was also speaking at ITB.

John hates filters. He thinks that forcing customers to use filters to narrow down their search for travel products is poor UX (User Experience). So he uses a specialist data analysis company, BD4Travel, to crunch numbers and develop AI solutions to make life easier for his 2 million customer base when they are perusing his 300,000 hotels in 1,000 destinations served by 200 airlines.

In a survey that TravelRepublic will be publishing this week, 17% of their respondents said although they wanted to travel more, they did not book a holiday because they were confused by the number of choices they were presented with. "That is phenomenally frightening!" says Musk, and it is what is driving him to use ever more sophisticated AI to improve UX.

Andy Owen Jones, CEO at BD4Travel says it's all about big data. They have tracked something like 600 million leisure travellers (anonymously) through their clients' websites (including tens of millions through Travel Republic) to determine 'inference'.

"We can see how if someone follows a particular pattern we can start to infer what it might mean for them" he says. It's way more reliable than the traditional customer database approach which says 'this person has booked three holidays to Cyprus with us, therefore he is most likely to be looking for a holiday to Cyprus'.

Is the step change in AI and its adoption just about more intuitive travel booking?

No, it'll soon affect us everywhere, especially where AI is used for speech recognition and interaction. Consider this conversation...

"You know how you like biltong?"

"Don't deny it. The chutney beef biltong, sliced, that you developed a taste for when you were in South Africa 19 months ago? You told me you liked it, and you've bought it twice since the Africa trip.

"Anyway, I've bought 500 grams for you. It'll be delivered on Thursday when your calendar says you'll be working from home.

"That's ok, you can thank me later"

Your fridge has always leaned towards conversational irony!

Welcome to the rapidly developing world of near-sentient AI, in this case being hypothetically demonstrated by your future fridge who you have authorised to make occasional surprise purchases up to a ceiling of £20. Trawling the Internet, it has spotted your favourite biltong selling for under £3.40 per 100g, the last price you paid.

Computers are already writing stories for major newspapers, helping doctors to search for cures for cancer, and winning game shows. In short, AI has crossed the chasm from science fiction to science fact, and according to PHD's book 'Sentience: The Coming AI Revolution and the Implications for Marketing' "within just over a decade from now, given the immutable march of Moore's Law - the mega trend in technology that says computers will progressively get twice as fast and cost half as much every 12–18 months - AI machines will be far more intelligent than we are."

The book's authors suggest machines, while not actually sentient, will certainly appear sentient, and over time we'll develop relationships with them to the extent that we'll allow them to make some decisions for us... to act as an agent, a Virtual Personal Assistant (VPA).

You can see where this is going.

In the travel and tourism sector there will be a conjoining of AI systems. Not only will AI infer what holiday components you are likely to book, but also, if you allow your personal AI device, it will be able to agree with the OTA's selection and book it! Before too long, say the authors of 'Sentience: The Coming AI Revolution and the Implications for Marketing' machines will be marketing to machines!

... if we let them.


(Image: Bigstock/Mike Kiev)