The future of machine learning: how are organisations preparing for the next big thing in big data?

Danny Palmer investigates how organisations can best benefit from machine learning - and how they plan to hire those who can help them prosper from it

Machine learning is a big deal and some of the biggest names in enterprise technology are doing battle over it.

It isn't a new technology - it is actually almost four decades old - but now using it isn't restricted to supercomputers, thanks to ongoing advancements and the manner in which some of the big players are attempting to commercialise it.

Indeed, earlier this year, Amazon Web Services (AWS) launched an artificial intelligence powered predictive-analytics service to help organisations improve the quality of their data-driven decision making.

The move represents Amazon following the likes of Microsoft and Google into the machine learning space as vendors look to take advantage of what they anticipate will be a boom for the technology.

But even now, machine learning isn't hiding in the background; it's at the forefront of how many organisations do day-to-day business. For example, it plays a critical role in Amazon's retail services, ranging from price comparisons with rival retailers in its online store to features within Kindle. But how is machine learning actually defined?

Gartner's ‘Hype Cycle For Smart Machines' report - which suggests the technology will be very common within five years - describes machine learning as "a discipline that provides computers with the ability to learn from data without being explicitly programmed".

Gartner argues that machine learning will have an "amazing" business impact in the next five years. But many businesses aren't sitting around doing nothing about machine learning, they are already thinking about how they can best leverage it to improve their operations.

One of those is TripAdvisor, the world's largest travel site, which already relies on data and algorithms to ensure that the crowdsourced reviews of hotels, bars, restaurants and other leisure destinations are accurate and free from fraud.

Jeff Palmucci, director of machine intelligence at TripAdvisor, describes information as the "main product" of TripAdvisor and that machine learning can play a big role in analysing and managing it all.

"Being able to organise that information and make it consumable by your users is a huge job of machine learning," he told Computing at the recent Yandex Data Factory "Machine Learning and Big Data: Business Challenges" conference in Berlin. "Whether there's a pattern in the data you can find and exploit for business purposes, machine learning is a good place to do that."

Manning the machines

However, there are hurdles to overcome in the race to exploit machine learning. One of those, Palmucci explained, is "finding good people".

"It's becoming a more trendy topic so a lot more people are paying attention to it, but there's some pretty heavy maths under there so you need the skills," he said, adding that hiring people with the right skills as "very tough".

"The advent of big data and everyone keeping track of it means the business opportunity for making use of it is huge, so everyone is trying to hire the same people," he explained.

Global marketing and research firm GfK also sees a big future in machine learning, especially when it comes to the area of predictive analytics techniques, as Norbert Wirth, global dead of data science, explained.

"I think what we'll see is we'll move more and more into the area of prescriptive analytics where the output is not a number, the output is an actual recommendation, an actual statement that is easier to process," he told Computing.

According to Wirth, machine learning won't be restricted to just crunching numbers and giving yes or no answers; it will be able to analyse figures to make specific recommendations as to how companies should act when making business decisions.

"Rather than saying the expected market share for your brand will give you these measures about this per cent, the output would be 'increase your TV advert spending by X to get this," Wirth explained, describing how he believes this area of "prescriptive" analytics is "going to be a big field" for GfK and others.

"Data is everywhere and for businesses and all industries, it's going to be really key in the near future to understand how we can leverage that," he explained, adding that the challenge of "how you distinguish between useful and not so relevant data" represents "a very important first step".

'Killer skillsets'

However, much like TripAdvisor, GfK has to overcome something of a hurdle in order to find the right "killer skillset" to get the best out of machine learning.

"You need the nerd but you also need hybrid people who understand the algorithms who can also translate that and talk to a business person. That's going to be the killer skillset in the future, someone who's capable of translating between business problems and the data world, that's the key person," he said.

But while there are organisations that are already chasing the unicorn of using the power of advanced machine learning techniques, Gartner Fellow Stephen Prentice says the actual process of using advanced algorithms to make decisions is already a widespread practice - it's just not referred to as machine learning.

"We will start to see organisations realise we're doing a lot of this stuff already, we're just not calling it that. You can send a digital video to an app in the cloud and it will identify faces for you," he said, describing Facebook's image-recognition tool.

But Prentice told Computing that it won't be long before organisations want to use advanced algorithms to do more and that will result in significant growth and development.

"We'll start to see organisations starting to realise they can put systems together, and in the same way that apps took off, I think the application of algorithms is going to be significant," he explained.

Prentice even suggested that the skills problem won't be an issue, because the nature of machine learning will mean bypassing the need to hire those required to actually programme the algorithms as the machine will be able to make alterations by itself.

"What it means is that you will be able to very quickly create a very powerful system without having to have employed those people who do things you don't understand; you simply take advantage of it," he said.

However, Prentice suggested there's still a major challenge for true machine learning to overcome, and that's people.

"The crux will come down to, do you trust the machine? That's probably going to be the biggest stumbling block to the broader use of these very smart systems in business," he said.

While Prentice admits that some may find the idea of a machine taking decisions as "deeply unsettling" - after all, there are increasing and understandable fears among many workers that eventually a machine will do their job and they'll be out of work - he suggests the expansion of machine learning will actually bring benefits to employees.

"The reality is we can probably use the machines to get rid of all the tedious boring bits so we can focus on more value-added, and that has to be good for everyone," he concluded.