From spam filtering to production line optimisation, there has been a boom in the uptake and effectiveness of machine learning systems over recent months and years – and no field has seen a more significant and transformative increase than digital marketing. This article seeks to explain some of the most interesting examples of machine learning used in online contexts today, and provides commentary on how digital marketers should seek to accommodate – and capitalise upon – these ever-evolving innovations.
Machine learning is one of the most useful and widespread manifestations of artificial intelligence currently in-use in a commercial context. A machine learning system is an algorithm with the capability to independently optimise its own processes by analysing and acting upon the data generated by its own activities. This technology is already used across a huge range of web applications.
Facebook’s algorithm is constantly learning you; analysing your behaviour in order to personalise the content it delivers to your unique newsfeed. As you would expect, likes/reactions, link clicks, video plays, comments and shares – engagements – are factored into the newsfeed algorithms calculations. Perhaps more surprising is the fact that the time you spend idle on your newsfeed, reading or viewing content without actively engaging with it, is also included in the algorithm’s calculations. Keep scrolling past a certain type of content and you’ll see a little less of it in the future.
Insight – Post engagements and reach are intrinsically linked – but the inclusion of idle time in Facebook’s newsfeed algorithm reminds us that the intrinsic value of a Facebook post as a standalone unit of content also has an important role to play. If you share links as part of your Facebook marketing strategy, it may be worth your while to experiment with accompanying those links with relatively lengthy copy. Facebook wants to keep its users’ attention – it makes sense that they would reward the content posters who help them do exactly that, on-site.
2016 is proving to be a year of radical change for Twitter, with support for longer video clips and an increased character limit for rich posts numbering among a raft of updates which appear to strongly indicate a new focus on photos and videos for the social messenger app.
This June, Twitter founder Jack Dorsey announced yet another move signalling strong ambitions surrounding visual content: namely, the purchase of London-based machine learning experts, Magic Pony Technology. Writing on the official Twitter blog, Dorsey stated:
“Magic Pony’s team will be joining Twitter Cortex, a team of engineers, data scientists, and machine learning researchers dedicated to building a product in which people can easily find new experiences to share and participate in.
“Magic Pony’s technology – based on research by the team to create algorithms that can understand the features of imagery – will be used to enhance our strength in live and video and opens up a whole lot of exciting creative possibilities for Twitter.”
Insight – While the exact functionalities under development by Magic Pony Technology’s and Twitter are yet to emerge, a statement on the former’s website clarifies the direction of their work: “[…] we’re excited to announce that we’re joining forces with Twitter to use our technology to improve the visual experiences that are delivered across their apps.”
It seems we may be bound for a future in which Twitter delivers images to users’ newsfeeds based not just on the words and tags used to describe images and videos, but also the subject matter of the media, algorithmically diagnosed.
Earlier this week we reported on the landmark news, divulged by a senior Google employee, that 100% of search queries received by Google are now processed by the RankBrain machine learning system, with a high percentage of search rankings affected as a result. RankBrain forms an important part of the over-arching Google Search algorithm, Hummingbird.
It’s unclear exactly which factors RankBrain takes into account when weighing the efficacy of search results, but what we do know is that the system is ever-evolving, ever-learning and ever-seeking to serve results listings that better meet the user’s requirements.
Insight – If you want to rank the highest, be the best. Bill Gates was telling us ‘Content is King’ back in 1996, but in light of the growing power and awesome potential of machine learning, it really does feel like the websites with the best content – in terms of quality, depth and relevance – are finally set to eclipse cleverly SEO’d sites populated with inferior content on Google’s results pages. The age-old skills of link-building, meta data optimisation and keyword planning are still important, but search marketers should start spending more time on honing content relevancy and quality.
Insight – Some writing we read for its style and voice, some to appreciate or reject a viewpoint, and some to receive information. Usually a piece of writing will offer a combination of these facets, but in some cases the reader simply wants the hard facts – especially in news or sports reports. In these situations, journalism AIs are already capable of carrying out the task with little or no input from humans.
As reported by The Guardian in 2015, American AI firm Narrative Science predict machine learning systems like their own will be capable of writing 90% of our journalistic articles by 2030. Here’s an example of a sports report, written independently by Narrative Science’s machine:
“Tuesday was a great day for W Roberts, as the junior pitcher threw a perfect game to carry Virginia to a 2-0 victory over George Washington at Davenport Field.
“Twenty-seven Colonials came to the plate and the Virginia pitcher vanquished them all, pitching a perfect game. He struck out 10 batters while recording his momentous feat.
“Tom Gately came up short on the rubber for the Colonials, recording a loss. He went three innings, walked two, struck out one and allowed two runs. The Cavaliers went up for good in the fourth, scoring two runs on a fielder’s choice and a balk.”
Insight – But is this a good thing? In the interests of journalistic objectivity, our answer would be a tentative yes. Machines may lack the extraordinarily complex morality and character of a human writer, but at this point they also lack the biases and pre-conceptions. Of course, as machine learning systems continue to evolve and self-educate, there’s nothing to stop them from developing creeds of their own, complete with all the baggage, beauty and intricacy.