Artificial intelligence isn’t new, but it’s finally getting smart, thanks in part to a type of machine learning known as ‘deep learning’. But does deep learning really hold answers for the travel industry, is the question posed by EyeforTravel’s latest report.
In recent years, deep learning has taken a big step forward and made some major advances in image and pattern recognition, autonomous driving and speech recognition (aka natural language processing). Like the vast network of neurons in the human brain, the artificial neural network, that powers deep learning, is an interconnected group of nodes, but it’s not quite the human brain – yet.
“Neural nets are still simpler than biological counterparts,” explains Alex Hadwick, EyeforTravel’s Head of Research. “And today what it [deep learning] truly excels at is focusing on a single task, which is typically finding relationships and patterns in very large quantities of data,” he says.
Unsurprisingly, in the data rich travel environment that is a compelling enough reason to take note.
Report interviewee Amer Mohammed, Head of Digital Innovation at Stena Line, certainly sees the value. But according to Mohammed, “what we are doing now is artificial narrow intelligence, AI that’s specific to a certain task. We need to come up with mathematical models that can actually understand the world, not just fake understand it.”
In the meantime, however, some of the tasks that deep learning is already being used for in travel include pricing, language processing, image recognition, consumer analysis, and market modelling.
For Trainline, a rail ticket retailer, using AI technology to make predictions on likely shifts pricing is an interesting application, because this would directly benefit the customer. According to chief operating officer Mark Brooker, Trainline is strongly focused on product development, and is investing in technology that enables it to delve more deeply into its data. “We are looking at technologies like machine learning, and what sort of outcomes can be achieved from artificial intelligence, but always with a focus on improving the customer journey,” he stresses.
So, what’s driving this growth?
Among the reasons, says the report, are “a democratisation of computing power and access to information via the internet, alongside the development of specialised computer tools, like graphics processing units (GPUs) – partly developed for the video game industry, and also handy for Bitcoin miners”.
It continues: “Our connectedness, booking via telephone, surfing websites, ‘liking’ things on Facebook, is also amassing mind-blowing quantities of data, stored in the cloud”.
A big factor is that the cost to invest in technology has come down. As Mohammed puts it: “Seven years ago, to do the things we are doing today, you would have to invest $12million in a super computer. Now we can pay $6,000 for our algorithms to run for three minutes on Amazon.”
Solving intelligence, solving problems
The technology continues to advance rapidly to bring AI even closer to having the human touch. DeepMind, part of the Alphabet Group, is one company that is making eye-watering progress in “solving intelligence” through, it says, “long-term thinking, interdisciplinary collaboration with academia, along with the relentless energy and focus of the best tech start ups”.
Of course, there are still issues and bottlenecks to overcome for artificial neural networks to mirror the human brain. Image recognition technology, for example, illustrates that “to work, neural networks need masses of classified data and for their errors to be corrected by human hands”.
They also a greedy when it comes to IT requirements. While the human brain runs on the equivalent of around 20 watts, AlphaGo, the DeepMind AI that beat the top Go player in the world, required 50,000 times that. On this subject, in the report, Russian metasearch company Aviasales shares this view: “System resource is the only limit…and, we could be more productive by achieving [a] new level of computer performance”.
What is clear is that there are numerous opportunities for travel companies.
“Almost everything in travel has a huge number of variables as trip itineraries are complex with multiple decision points, making deep learning especially suited to drawing conclusions from the masses of data,” Hadwick explains.
In addition, as more data and information is added, the better AI gets, so it could be a truly powerful tool for personalisation.
Neural networks can end up being black boxes due to their complexity
But there could be ethical challenges. “Neural networks can end up being black boxes due to their complexity, and multiple layers of decision-making. So, you can find yourself with an answer but not knowing how the AI arrived at it, and this, for one, could run counter to European data regulations,” Hadwick warns, adding that there is a risk too of human bias and error creeping in.
That’s not a reason not to do it. International Data Corporation, after all, predicts that AI will drive worldwide revenues exceeding USD47 billion by 2020. And smart travel firms “which know where customers are, go and behave – could serve themselves a deep slice of that pie”.
The most successful travel brands will, it seems, benefit from AI in the future, but providing they apply themselves with intelligence.
Does Deep Learning Really Hold Answers? is part two of EyeforTravel’s How Will Artificial Intelligence Transform Travel? series. You can find the first report, which studies chatbots in travel, by clicking here