The evolution of machine learning technologies is changing the world around us, benefitting both businesses and consumers alike. Ideas that were once only figments of the imagination have now become a part of reality. As a form of artificial intelligence, machine learning ingests large amounts of data and adapts accordingly, meaning that it can only advance in the coming years. Although there are numerous examples of modern machine learning applications, we have focused on ten that most likely affect you on a daily basis.
1. Facebook: Improving the user experience
The Facebook user experience continues to improve thanks to machine learning. These technologies help predict actual clicks on the ads that show up during a user’s session. When it comes time for a user to find new “friends”, Facebook has plenty of recommendations.
One of the newest features on Facebook is DeepFace, which certainly has the potential to raiseprivacy concerns. Using Machine Learning, DeepFace can identify Facebook users in pictures. Perhaps you’ve already seen this technology in action – I know I have. Although DeepFace is capable of identifying a single person in any of the 400 million new photos uploaded daily across the site, Facebook has restricted it to only identify users to their friends.
2. American Express: Keeping customers’ best interest a priority
American Express uses machine learning for a variety of purposes, all performed with the best interest of its customers in mind.
To protect its customers from financial loss, American Express uses machine learning for fraud detection and prevention. The company leverages data sources including cardholder information, spending details, and merchant information to halt fraudulent transactions.
American Express also has a phone app that customizes restaurant suggestions by using arecommendation model. The machine learning app uses card member profile and recent spending history to train the model.
3. Netflix: Personalizing the viewing experience
Much like American Express, Netflix utilizes a recommendation system to personalize the viewer experience. The system relies on both explicit and implicit feedback from the user. Explicit feedback incudes ratings, reviews, and wish lists, whereas implicit feedback incudes purchase history and view time. Netflix, therefore, tracks what its users watch to make accurate suggestions for future viewings. This approach seems to be working for the company because, according to Netflix, 75% of what people watch comes from recommendations.
4. Amazon: Personalizing the browsing experience
Amazon also uses a recommendation system whose algorithm personalizes sessions for returning users. The recommendation system depends on information including a user’s purchase history, items added in the virtual shopping cart, items rated and liked, and what other customers have looked at and purchased.
5. Pandora: Optimizing the listening experience
Pandora, the popular music streaming service, combines machine learning and human coding to perfect the listening experience for its users. Each time an account holder uses Pandora,preference data is collected. This means when a user likes or dislikes a song, skips a song, or simply isn’t actually listening to Pandora - it is all tracked. The company’s employees code the music according to 450 values, such as tempo and the number of vocalists. According to the company’s chief scientist, this manual practice will remain in place for sometime because machines cannot understand music in the way that people do. Pandora’s system then combines the codification and data retrieved from users to determine which song to play (that the user will actually listen to!).
6. Google: Making strides towards a more effective drug discovery solution
Google is currently involved in a research project that, in time, will help create drug treatments for people suffering from certain diseases. The task of drug discovery often requires high-throughput screening processes, which are both time consuming and expensive tasks. Google has been experimenting with machine learning in virtual drug screening to either enhance or completely replace the alternative. Machine learning will eventually make the process more effective by incorporating data from multiple diseases, as opposed to one disease.
7. USPS: Creating an efficient, accurate mail delivery system
The United States Postal Service (USPS) specifically uses pattern recognition technology to quickly and accurately deliver mail. USPS facilities contain Remote Computer Readers (RCRs) that scan non-barcoded mail for address information. In the process, the information contained on the envelope or package is compared to Address Management System (AMS) databases that contain only addresses, no names. The mail is then given a barcode for processing. In the event that the RCR can’t find a match, a picture of the address is sent to a Remote Encoding Center (REC) where employees manually enter the address. The results are then sent to the facility with the physical piece of mail. RCRs have greatly benefitted the USPS by improving the efficiency and accuracy of mail delivery to residents and businesses in the U.S., allowing the company to save money. In fact, technology advancements mean fewer manual entries, allowing many REC facilities to close. There are only 15 facilities remaining today.
8. Target: Implementing best communications for brand reaction
Target uses machine learning to track every interaction a customer has with the brand – from mailings to website visits to physical store visits. The technology uses this data to discover how to best communicate with customers to encourage their reaction to the brand.
Target’s machine learning technologies are incredibly accurate at identifying changes in customers’ lives. You may remember the controversial news story about Target discovering a girl’s pregnancy before her own father. How did Target do it? The company’s statistician analyzedbasket data to determine pregnancy. Specific products then became a part of the analysis, allowing Target to create a pregnancy prediction score. This score indicated to whom Target should send these coupons.
9. Microsoft: Familiarizing video games with player behavior
Although always present in video games, machine learning has become more sophisticated as of recently. Technology giant Microsoft is incorporating this advanced technology into video games for insight. Microsoft studies the data from Halo3 to understand gamer performance at specific levels and to learn when players are using cheats. Microsoft also worked on Drivatar, which is in the game Forza Motorsport. When the player initially starts, the game does not know his/her driving style. Over time, beginning with practice laps, the game gets familiarized with the player’s driving style. This is crucial to entering new game levels and to playing with others, including friends.
10. Google News: Determining and ranking the news of the day
Computer algorithms determine the top news of the day on Google News, meaning there is no human intervention. These algorithms take different factors into consideration for news judgment including how frequently a story appears online and where it shows up. Essentially, Google News relies on the news judgments of other news organizations for prominent stories that belong on the home page. In terms of identifying related stories, Google implements clustering algorithms. This looks at factors such as title, text, and publication time.
The use cases referenced reveal that machine learning technologies have greatly improved people’s lives, perhaps even yours. It’s typically a win-win situation for both the company and the customer. While the company possesses the technology to predict future behaviors, the customer is provided with a personalized experience. At the current rate of growth for machine learning, one can only imagine the possibilities in the coming years!