Beyond the Buzzword: Testing Your Machine Learning Knowledge

Amberly Dressler
by Amberly Dressler 21 Feb, 2017
Beyond the Buzzword: Testing Your Machine Learning Knowledge
Online searches for "machine learning" are at their peak according to Google Trends with search interest doubling from Jan. 2016 to Jan. 2017 - despite many companies using this method of data analysis for years.

SDL uses machine learning, for instance, so that if its translation software interprets a word or phrase incorrectly - and it's corrected or over-written by a human translator - the software solution automatically fixes the mistake and does not repeat it. Similarly, Google algorithms already learn from a person's searches to deliver more personalized (better) results over time. So why the sudden interest in machine learning (where systems essentially "evolve" when new data is introduced)?

The popularity is driven in part because humans are, well, flawed. People don't always learn from their mistakes or act on data in the predictable way a computer would. Thanks to the ability for machine learning and artificial intelligence (AI) initiatives to work in tandem, however, that is starting to change and creating some rather valuable opportunities for enterprises in the process. For instance, financial services company USAA tracks a customer's behavior and looks for patterns that don't match in order to catch identity theft. By learning from member data, USAA improves its chances of identifying malicious behavior and could, essentially, carry out entire conversations with customers regarding suspicious activity via chat bots (the AI component).

The possibilities are endless. Let's explore the current buzz around machine learning, in particular, and some of its current applications in this month's Quiz Time. (Note: answers are at the bottom of the page.)

1. True or False: Facebook uses machine learning as part of its "On This Day" memories feature by learning from the user (whether they shared similar memories or dismissed them), understanding visual concepts (identifying what is in the photo and where it was taken), and even the automatic filtering of memories that were previously dismissed or are with exes or people who are blocked.

2. Enterprises can use "predictive analytics" offerings from vendors including Amazon and its machine learning system to create a wide variety of applications including:

a. Applications that flag suspicious transactions, detect fraudulent orders and forecast demand
b. Apps that personalize content, predict user activity and filter reviews
c. Apps that listen to social media, analyze free text and recommend items

d. All of the above

3. Which retail company openly promotes its use of machine learning and has an official technology blog in which data science is a regular topic?

a. Stitch Fix
b. Gap
c. Target

d. Kate Spade

4. Which is true of Waze, the popular traffic app that learns from itself and its users?

a. Using past and real-time data, Waze predicts traffic jams and navigates users around them as well as calculates and reminds users of the best times to leave for trips - getting smarter as more data is collected from app users
b. Google purchased Waze for roughly $1 billion in 2013
c. The app was proved vulnerable last year in that hackers could create "ghost drivers" to track users' where-abouts in real-time and use the same bots to resemble traffic jams within the app (somewhat resembling a real-life DDoS attack)
d. All of the above 

1.  True; 2. d; 3. a; 4. d
Amberly Dressler
Amberly Dressler
Head of analyst relations, public relations, customer advocacy (People Heroes), customer community, content marketing (full funnel/lifecycle), content operations and optimization, reputation management and social media. Leads a team of nine superstars to exceed our goals multi-fold.