Impact Publications : MICEBTN-67-JUNE
Page 30 • MICEBTN - JUNE-AUGUST 2017 Artificial Intelligence called the ‘new electricity’ travel AI is coming whether we want it or not – and here’s why THERE’s an old joke that goes: “I’ll believe in Artificial Intelligence (AI) when there’s a computer that un- derstands the difference between ‘Concorde flies like an arrow’ and ‘fruit flies like a banana’.” AI stems from the 1980s, when some IT gurus claimed thought-ca - pable computers were just around the corner, writes Jack Handley. They lied. Roll on 30 years or so, however, and the ballpark is so different that Chinese tech company Baidu’s sci- entists call AI “the new electricity”. Travel AI is complex and the stakes are huge. Flight disruptions alone cost the aviation industry an estimated US$25 billion a year, according to IT firm SITA. Mislaid bags add to that total, while lost sales opportunities covering seat and meal selections and excess baggage charges are worth many billions more. As a result, airlines today are investing hundreds of millions to boost their on-line ancil- lary sales. And it seems to be worth it. Lufthansa already analyses cus- tomer behaviour to decide when to offer individual customers tar- geted additional services, such as an upgrade from economy to busi- ness class. The carrier says twice the number of customers took up upgrade offers after the targeted system was implemented. In addition, we’re already seeing smart tools that enable ‘conversa- tions’ with voice-enabled ‘chatbots’ that have encyclopaedic knowl- edge about company travel policy, destinations, best fares, hotel rates and so on. Other developments mean that say a bag is delayed (it goes astray but the airline knows where it is), a chatbot can message the ‘victim’ and tell him not to wait at the bag- gage carousel, ask him where he’ll be staying that night and promise to deliver the bag there. Other ideas include having bots contact passengers at airport check-in to offer a deal for one- time access to a business lounge or a seat upgrade. Knowing the customer Travel AI works by taking infor- mation about what individual trav- ellers already do (they buy airline tickets, hotel rooms, cruises and tours, visit attractions, restaurants and shops) and combining that information with their preferences (left side aircraft window seats, non-smoking hotel floors, outside ship cabins with balconies, all-inclu- sive luxury coach tours, zoos and wildlife sanctuaries, vegetarian res- taurants, white wine and upmarket souvenirs) ... you get the picture. The difference between say 2010 and now is that today’s databases can collaborate with others to build increasingly comprehensive AI sys- tems. But AI needs huge amounts of data and consumes more num- ber-crunching power than entire data centres did just a few years ago. So behind the scenes, a new class of computer processor, the graph- ics processing unit (GPU) is replac- ing the old central processing unit (CPU) used by most desktops and laptops today. Firms including Google and Mi- crosoft now are buying GPUs from companies such as Nvidia - or are designing their own. For comparison, consider that GPUs have hundreds of specialised ‘cores’ (brains) all working in paral- lel, whereas slower CPUs have only a few cores that tackle computing tasks sequentially. For example, Nvidia’s latest GPU processors by Jack Handley Data centres such as this one are moving from CPUs to GPUs to handle the industry’s growing demand for Artificial Intelligence capability.