Machine Learning market is growing: from $1.29 billion in 2016 to $39.98 billion by 2025. What has contributed to this growth? Machine Learning was born in 1959, coined by computer scientist Arthur Samuel, but only recently has the larger business community come to understanding of its value. In the nearest future, it will be used by everyone from the most successful companies to small private companies. We have compiled a list of companies from various fields that use Machine Learning. How many of these companies do you know?
Spotify: «Music Recommendations»
Every Monday the music service Spotify provides each user with individual playlist of 30 songs that you like is Discover Weekly. The recommender system Discover Weekly was presented in 2015. The system works by dint of three types of models:
Collaborative filtering model which compares your music preferences with the preferences of other users
Natural Language Processing model which works on the basis of text analysis
Audio model which analyzes raw audio
Spotify has become the most popular music streaming service in the world thanks to such a powerful recommender system. No matter how seldom you listen to music, and whatever it may be, the algorithms will advise you something similar.
Netflix: «Personalized Film and Cover Recommendations»
The main feature of Netflix is the personalization recommender system where each user sees on the main page those films and TV shows that may interest him. Also Netflix shares information on the Medium about how, through of Machine Learning, makes content more attractive by changing the covers of movies and TV shows.
Previously, on Netflix for each film there was one cover, which was chosen based on the interests of the audience, this is the one that was often clicked. Different users are shown different illustrations which meet their preferences. For each film and TV Show is displayed not a fixed image, but the one that directly depends on the films that he has already watched. For example, if the viewer is a fan of comedies, then the cover will show a funny situation or a famous comedian either Jim Carrey, Adam Sandler or Eddie Murphy.
The process of selecting illustrations for films is completely automatic and is based on data collected about each user of the service.
LinkedIn «Resume and Content control of News feeds»
The social network LinkedIn in February 2018 launched an AI-assistant, which helps applicants to compile a resume. An assistant selects similar questionnaires in the required area and gives a piece of advice. AI also suggests using different terms to make the document more informative and interesting for employers.
LinkedIn uses a combination of machine learning algorithms with the work of the human operator. This allows you to classify content by quality. ML works like a trigger, evaluating content in two stages. These are:
LinkedIn classifiers working in online mode mark each image and text like a "spam", "low quality" or "good" in real time
LinkedIn classifiers that run as you read the content determine the likelihood of users' interest and the quality of this post
Salesforce: «Data mining in CRM systems, Text extraction and Chatbot-polyglot».
Salesforce is an American developer of the CRM system of the same name, which is provided to customers exclusively on the basis of the SaaS model (Software as a Service). In 2016, the company launched integration with Einstein AI and added new features to the system in 2017. Einstein is an assistant based on AI for data analysis in the field of CRM. Its main function is to help companies identify, forecast, recommend and automate the most effective business processes. The assistant uses Data Mining and ML to for forecasting future sales figures of the company.
Also in 2016, Salesforce bought a startup MetaMind, specializing in Machine Learning and Natural Language Processing. Developed MetaMind's algorithm based on neural networks analyzes the English text and makes of it a brief excerpt in the form of a retelling by introducing new words and expressions. According to journalists from the MIT Technology Review, the algorithm Salesforce's produces "surprisingly connected and accurate" extracts: «The news from The New York Times of 345 words has been transformed into three sentences of 50 words».
In 2017, a research team at Salesforce came up with a clever way to improve the performance of many modern language programs. One of it is teaching an algorithm to speak another language before training it to do other tasks. The translation system works only between English and German. But the developers promise to add support for other languages.
Evernote Corporation: «Automation of the Workflow of Users»
Evernote Corporation in 2018 expanded its self-titled app that is used to create and store notes, with the capabilities of artificial intelligence. Tools added to the Spaces function.
AI tasks are improving search functions and performing popular tasks of users. For example, artificial intelligence scans the contents of notes and reminds you to perform a task.
Since that case, the universal algorithms could not demonstrate the correct operation; the developers used AI, which is adjusted for a unique user.
Spaces functions are available only for corporate users. But in the future the company is planning to add new features of the application to personal accounts.
To be continued