Five Ways You’re Already Using Machine Learning: A Day with AI

Print Friendly, PDF & Email

In this special guest feature, Mark Scott, CMO at Apixio, highlights the prevalence of machine learning in everyday life and offers five ways you’re (probably) already using machine learning all without you realizing or thinking about it. Mark has more than 19 years of medical technology and health care provider marketing experience. His expertise covers all the bases—from brand development, positioning and messaging; to brand identity, packaging and labeling; public relations; content marketing, website development; internal/employee communications; and global brand-launch activations. Mark has a Bachelors and a Masters Degree from the University of Western Ontario.

“Machine learning” can seem like a scary term, bringing to mind images of the techno-dystopias portrayed in the Matrix, Terminator, and Black Mirror. However, far from these dark narratives, machine learning has been reaching ordinary people for a long time in simple, common, and helpful ways. Here are five such ways you’re (probably) already using machine learning.

1. Checking the situation on Google Maps

Most of us, especially those that drive, have used Google Maps before. However, not many ask: How’s this nifty thing done? How does Google construct maps that reflect every street, avenue, and alley with such accuracy? The answer is artificial intelligence, or AI, and machine learning to be specific.

Google Maps is fueled by accurate real-time information from thousands of images of street signs and locations, which are analyzed by Optical Character Recognition algorithms. The task is challenging because the algorithms must know to ignore extraneous information — like the lettered shirt of the person standing next to a street sign for instance — and to pick out just the relevant information. To accomplish this tough task, Google trained its machine learning algorithms using the difficult French Street Name Signs (FSNS) database. Once the algorithm could correctly recognize these complex images, it grew much more accurate.

2. Ordering a Lyft or Uber

Whether you’re team Lyft or prefer to hop in an Uber, both services, much like Google Maps, power decisions with AI. Driver assignments, driver ETA, and your ETA at your final destination – are all calculated by algorithms that are constantly tested and refined in real time, using machine learning and the massive quantities of data from drivers and customers.

One important thing for many; rideshare companies are using machine learning to help beat the dreaded ‘surge price’. Surge pricing, or time-limited price hikes, currently compensates for times when there are not enough cars on the road to supply all the passengers who want rides. Ideally, machine learning could anticipate times of high demand (say, commute times in April on the East Coast when it frequently rains) and incent cars to be on the road, in advance.

3. Using spam filters and priority tags to keep you organized

Your inbox probably seems an unlikely place for machine learning, but AI technologies are in fact the engine behind the “Spam Filter”, one of email’s most important tools.

Simple rules-based filters are used for the spam filter. Think of the words and phrases “pharmacy”, “you’ve won the lottery” or “Nigerian prince”. While you may very well be friends with a Nigerian prince, if the message seems suspicious, and is coming from an unknown sender, then it will probably get flagged and kicked to spam. The filter ‘learns’ the content of the emails, identifying signals by gathering inferences from word relationships. The filter is able to counter-maneuver spammers that try to outsmart it with updated messages.

In addition to this general work, the filter also ‘machine learns’ what you, the user personally consider spam. It does so by ingesting data on what you delete and mark as spam mail. This data works in conjunction with data contributed by the entire user base. With these approaches working in tandem, some reports place Gmail’s spam filters at a 99.9% success rate. Additionally, researchers tested the effectiveness of Priority Inbox on Google employees and found that those with Priority Inbox “spent 6% less time reading email overall, and 13% less time reading unimportant email.”

A similar approach is used to tag emails for predetermined primary, social, and promotion inboxes, as well as automatically labeling emails as important.

4. Preventing credit fraud

A database of consumer complaints by the Federal Trade Commission reports 1.3 million (42%) of the 3 million complaints in 2016 were fraud related, with losses of $744 million. A total of 55% of all cases were fraud or identity theft related – that’s more than 1 in every 200 Americans per year. Keep in mind, 7% of American households are unbanked, and only 77% of Americans are over the age of 18.

Big problem, right? But how do financial institutions even determine if a transaction is fraudulent? Bank of America alone has an estimated 58 million customers and like most banks, its daily transaction volumes are much too high for manual review.

To sift through this mass of data and distinguish normal purchases from illegal ones, banks use machine learning. For example, FICO, well-known producer of credit ratings, uses neural networks to predict fraudulent transactions. Named for and meant to simulate the neural networks in our own brains, these systems analyze examples of labelled data, developing their own characteristic markers, and learning how to identify unlabelled inputs. Factors considered include: the customer’s recent frequency of transactions, transaction size, and the kind of retailer involved. (It’s good for the bank too— MIT researchers found in 2010 that machine learning applied to customer transactions, could reduce bank losses from delinquency between 6 and 25%).

5. Posting on Facebook

Facebook uses machine learning to ensure that you see the most relevant news from your friends. Many of us have hundreds of Facebook friends, not all of whom are equally close to us. Sociologists have actually determined that it’s implausible to have more than 150 close friends — it’s called the Dunbar number. Machine learning helps Facebook determine which one of your connections is truly a close friend whose life you want to hear about, and who is your cousin’s ex-boyfriend who showed up at Thanksgiving once.

Facebook goes through all potential posts you could see, and assigns each one of them a score. The score is determined by a couple things; your previous “like” and “comment” activity; whether the post is an interaction between two people or between a person and a corporate page; whether the post is from one of your friends or a friend of a friend; whether the post has seen a lot of time-consuming activity (many people commenting) or is merely “liked”.

The scores of all your potential posts are ranked, and that ranking becomes your Newsfeed.

Just the tip of the iceberg

All these examples are just the tip of the iceberg, AI is being used in so many different fields – anything from fashion, where companies like Stitch Fix use machine learning, customer feedback and stylist expertise to deliver clothes to your doorstep by subscription, and collect over $730 million in annual sales; to sports coaching,where it’s applied to thousands of historical strategies and plays to make predictions and recommendations based on this training data and user inputs.

Machine learning helps you throughout your day— all without you realizing or thinking about it.


Sign up for the free insideBIGDATA newsletter.

Speak Your Mind