# Machine learning applications: top 10 well-known companies and startups using the ML technology to improve their business

You’ve probably heard of Pinterest before. This is a platform where users can place and promote their content. In Pinterest’s case, the content is images, or arts. It’s one of a kind platform for being an artist on the internet - popular pictures from Pinterest tend to go viral. An improved feed could help users quickly find what interests them. Maybe that’s why this company decided to invest in machine learning for their web and mobile apps.

In 2015, they acquired a machine learning company Kosei, that uses ML technology for commercial applications. It’s specifically focused on algorithms of content search and recommendations. Machine learning algorithms provide users with accurate recommendations. Then based on their reaction ML technology predicts whether the content will be pinned by the user. Kosei also predicts the effectiveness and relevance of personalized ads using machine learning algorithms in collecting customer experience. This process takes place through data analysis, as well as the collection and discussion of customers' experience. Machine learning is trained with Spark, which improves its efficiency in real time. Thus, it improves the search for content in the application, giving the user the most accurate and suitable result. So now Pinterest is using machine learning for almost everything in their application - from advertising to moderation. They reduce spam and inappropriate content in the app. Isn’t this what made them so popular all over the world?

As you can see in this graphic, the international popularity of Pinterest has experienced significant growth since 2016 to the first quarter of 2020. Today Pinterest has up to 367 million users worldwide. You can easily connect one thing to another, and notice the results. Commercial algorithms for content discovery, based on machine learning, made Pinterest one of top-5 applications among US users. With the smart algorithmic feed Pinterest is now able to improve the control over user’s activity and helps everyone find exactly what they need right now.

Oh, Twitter, a place of every argument on the internet. This sweet world of something funny and revolutionary at the same time; it is Twitter where you can find everything from a dumb joke to political speech. And not so long ago Twitter had changed its feed to make it algorithmic.

Machine learning algorithms find the “most interesting tweets for you” and show them at the top of the feed. They analyze tweets in real-time and score them from “the best” to “the worst”. Maybe you don’t like such a change and you’d like to see your feed in chronological order, but don’t be pissed off just yet. Give Twitter a chance!

These algorithms make a decision based on your personal interests and preferences, and whether you like it or not let’s face the truth - you won’t quit your Twitter account anyway.

But there are good and less frustrating examples of ML technology in Twitter application. For example, it moderates the racist and other inappropriate content in your feed. Along with that, machine learning helps Twitter fight terrorism! In the past, ML technology helped to take down more than 300,000 terrorist accounts.

In the case of Twitter, Machine Learning technology might not be used in the best way possible and had driven application users to give negative feedback about it. So, the one thing you must remember - if you want to implement machine learning into your application and don’t yet know what positive changes it’ll bring to you, you’d better be as popular as Twitter. If not - think twice (or get help from your technical partner).

Facebook using AI for a number of features that help them to stay popular for so many years. We can say without a doubt that social media is inclusive in a good way.

The first thing they use machine learning for in their application is chatbots. This means that every business account on Facebook, with the help of ML technologies, can create itself a chatbot for better customer service. It isn’t, of course, a complicated software, but it can be extremely helpful for pre-seed/seed startups with low funding. A chatbot has to be created by the startup's developers themselves. However, they can add it to Facebook Messenger and program it for any necessary activity - whether it is statistical data for forecasting using ML algorithms or a unique adaptation to each user with the use of AI technology.

The second thing is the algorithms for reading the text for visually impaired people. This helps people who can’t see the text or a picture to hear it.

And the last is something similar to Pinterest ML - Facebook uses this technology for spam moderation and content filtering.

Facebook’s ML workflow is based on users’ data analysis to provide clients with the best results. For example, if its goal is to analyze user experience, it will collect data about the customers’ activity in the application, and then display an indicator of the customer satisfaction level.

As you can see on the chart above, Facebook keeps on growing and gets more users every year. In 4th quarter of 2019 this social media gained 2.5 billions users (most of which are located in Asia), and who knows what comes next.

We’re not saying that all the credit for this goes to machine learning algorithms in their app, but it definitely earned a place in Facebook success. This is achieved in the same way as in the example above. Facebook ML algorithms analyze content by key parameters and learn based on sorted publications.

Google's ambitions are so endless that it is easier to name something that Google isn’t working on than write a list of everything they do. But here we want to know only stuff related to AI and machine learning, so let’s say a few words the way about how Google is using this technology.

The most exciting thing Google did in this industry is the neural network. Maybe they’ve done a step towards the dystopia that Matrix, Terminator, and others were warning us about when they’ve built the machine that dreams. Even though DeepMind is more like of a machine that memes.

This has driven Google to a new wave of popularity because a neural network that uses other pictures to draw made a lot of masterpieces like the one above. By the way, this is not everything Google can do. Remember when you search for something, Google suggests the ending of your sentence? It is also possible thanks to machine learning. As well as those incredible things that Google does when fixes your misspelled words.

Also, Google Translator can learn and adapt to your writing style if you’re using it long enough. Thus, Google’s ML adapts to the most frequent requests of the user and to the unique pattern of the user’s actions. How is the ML process at Google? Depends on what this process is aimed at. If this aims to create a picture, as in the example above, then the technology will collect data and format it in the correct order to create the image. If the goal is to find the right route for your trip, Google Maps uses data about your location, analyzes it with machine learning algorithms, and provides you with information about vehicles or route that suits you best. All this is done in order to improve the customer's experience and provide search results that are most relevant for a particular user.

Baidu is a popular Chinese search engine that took an example from Google. They’re investing in AI applications as well, but they are aiming elsewhere. If Google’s neural net creates pieces of art, Baidu’s technology Deep Voice can even create a human voice.

Well, not quite human but at least it sounds very much alike.

In this case, an example of machine learning technology looks something like this: there is a collection of samples and technology performs their qualitative analysis. Hence, it finds info on the accents and specific pronunciation of the human speech depending on different countries and regions.

This neural net can imitate intonations, accents, and pronunciations to create sounds that exceptionally accurate. It can have a major impact on voice search since technology is learning by analyzing human speech. Machine learning applications of this kind also can be used for live translations or for transforming voice conversations into text. So, as you can see ML technology does have a potential.

This company’s AI Watson shows how fast an old company can adapt the rules of the new world. Watson is famous for giving well-formed recommendations on how to cure cancer, and hence this technology is slowly but effectively changing the world. This machine learning-based technology can predict scenarios of every individual patient curing process and find the most efficient treatment for each one.

IBM also has a mobile application for people with diabetes, which shows great results in the healthcare industry. For example, in 2019 the app predicted hypoglycemic events with 90% accuracy. By the way, if you’re thinking about implementing an ML solution from IBM into your business - we can help you! MassMedia Group is using IBM Watson's speech to text technology for high precision speech analysis. This can help you optimize your company’s work with text. In this case, different speakers are recognized in the audio materials with the addition of these key points in real time with high accuracy and confidence.

The companies we’ve mentioned above are well-known and had their peak of popularity long before today. But machine learning is popular not only among industry giants. There are successful ML using startups and now we’ll talk more about their cases.