Machine learning: what is it and what is its relationship with the ai?

Today we want to teach you in more depth one of the terms that has revolutionized and will revolutionize some interactions as we know them. We are talking about Artificial Intelligence and its most specific branch, Machine Learning or Automatic Learning.

As you may know, computing is always in constant evolution and what we can buy is usually not as cutting edge as possible.

For example, while we are developing the 4th generation of PCI-Express , researchers are already developing PCIe Gen 5 and studying the jump to the 6th . For this same reason it is not uncommon to find technologies that we did not know doing tasks that we had never heard of.

But before we go any further, let's narrow down the topic that we are going to talk about because, what is Machine Learning ?

Index of contents

What is Machine Learning ?

Machine Learning is a specific branch of computer science and Artificial Intelligence where systems capable of automatic learning are created .

This branch began its study and development around the 80s and today it is quite developed. For this same reason, both Artificial Intelligence and Machine Learning are used in many scientific and everyday fields.

In this branch, AIs are made up of one or more algorithms capable of processing large amounts of data and learning accordingly. The two key ideas on which this topic orbits are:

• The system must be able to analyze data and build skills that it did not have at its birth. Intelligence must be able to do the work autonomously, that is, without human supervision.

In the real world we have practical examples such as the classification of spam in emails, related recommendations on Amazon or predictions of the future using company data. The latter is an interesting section that more and more companies are betting on.

Using Machine Learning we can see what patterns identify dissatisfied customers or ex-customers to try to improve the relationship with other users in the same state. Seniority, number of complaints, contracted plans and others are studied to create certain profiles. Once the AI's conclusions are drawn, a group of marketing experts can create a specific campaign to combat those problems.

Thus, the company can create plans to attract or keep customers based on certain assumptions and goes from a reactive strategy to a proactive one. It is a very interesting tactic that uses Artificial Intelligence , large amounts of data and Machine Learning .

How are Artificial Intelligence trained?

For an Artificial Intelligence to be prepared it has to go through different phases:

1. It goes through a controlled environment first. Here you enter a large amount of data and their respective results with which you can create relationships between ideas. This part is called Supervised Learning . Then you are put into a free and unanswered environment where the AI itself will have to select a result. By knowing if your answers are correct or not, you create new rules in your algorithm. This stage is called Unsupervised Learning . Finally, an environment is prepared for him where he falters. If, for example, it is difficult for you to differentiate images with low luminosity, perhaps you are trained with night photos. This phase is called Reinforcement Learning. The process can be done from step 2 as many times as you want to fine-tune the Intelligence .

Generalized scheme on Machine Learning

A practical example would be to show an AI ten million photos and tell them which are dogs and which are not. Here he will relate that dogs usually have fur, they usually go on four legs and there are different shapes and sizes depending on the breed.

Afterward, he is given a million photos to classify. Here you must answer whether or not there is a dog in the photo and according to whether or not you will create new 'ideas' in your database. To implement this new data, Intelligence will establish new rules in its algorithm and now, for example, it will be able to differentiate dogs from cats.

Finally, his efficiency is studied and new photos are prepared to train his weak points.

Of course, this is a simple and very repeated system for the demonstration, but there are other more experimental and peculiar methods.

Tay, the Twitter bot

A recent case of experimental training was Tay , an AI developed by Microsoft designed to learn to express itself as a human.

Tay's Twitter Profile

The bot was programmed to initially speak as a 19-year-old girl and on March 23, 2016, she was released in the dark places of Twitter.

You were programmed to speak to the community and learn from the messages you received as well as your interactions with users. Her learning was almost completely autonomous, although she had to be withdrawn after 16 hours for showing negative behaviors.

In the short duration of his life, he tweeted more than 96, 000 tweets. However, the intentional offensive behavior of this social network made it quicker than soon for Tay to respond with racist and other phrases.

In this case, the Supervised Learning and the series of basic rules should have been duly revised. Knowing the carefree and offensive tone of the social network, Tay was not prepared to differentiate the real from the sarcastic. For the same reason, some users managed to easily "break" the "intellectual barrier" of Intelligence .

Machine Learning applications in the real world

We have already told you about some daily uses that perhaps you already knew about Machine Learning , but what other cases exist.

Below you will see a series of practical applications of this technology in the most common problems. Of course, they are cutting-edge solutions, so they also usually require considerably more money.

Health

A technology for a new type of clothing capable of reading information about our body is under study. It may be able to read our pulse, breathing, or anxiety.

These data are read by an Intelligence that evaluates the state of the patient in real time. So if you have a problem such as a heart attack at a specific time, you can diagnose and / or respond more quickly.

On the other hand, some bots capable of detecting suicidal thoughts have been implemented in some people. The famous Facebook Intelligence reads conversations and your activity to recognize patterns of suicidal tendencies, although there are other versions that study more closely the behavior of the person, his tone of voice and his body language.

Finance

In economics, some banks and companies have used Machine Learning- based solutions to detect and prevent fraud.

On the other hand, something similar is also used to more easily identify investment opportunities. It is also used to decide when to sell or buy stocks and other means.

Marketing

This we have already mentioned, but it is one of its best known applications.

It will have happened to you to see a couple of products on Amazon , enter Facebook, Google or Instagram and see just that product in your ads. It is no coincidence, since social networks and Google implement Intelligences that study your history and your possible interests to capture them where they can.

Some users see it as an intrusive way of 'attacking' the user and it is not surprising since they bombard you with an idea. However, advertising will move in that direction as it is more personal and the ads will be targeted to potential buyers.

Machine Learning and Deep Learning

These two terms usually go hand in hand, but they are not exactly the same. In a future articles we will talk about this second term, since it is something that deserves to be learned.

WE RECOMMEND YOU How to uninstall AMD drivers cleanly and easily

In general, we could establish the relationship between Machine Learning and Deep Learning as the one that Artificial Intelligence and Machine Learning have . Deep Learning is an even more specific branch of Machine Learning .

It shares key sections such as evolution over time and experience, but it has another series of differences.

Simplified Deep Learning

Its basis for learning and processing data is to use different layers that act as if they were neurons. Therefore, we could establish that these Intelligences are usually more refined, but also more complicated and expensive to build.

Although if you are more interested in this topic, stay tuned to the website and visit our next article on Deep Learning .

How far are we from Skynet ?

We have this section for the most dreamy minds.

This is a very repeated topic in books, movies and others. Not for nothing there is exactly a genre or theme called Cyberpunk . However, far from those futuristic dystopias controlled by Artificial Intelligence , our machines still have a long way to go.

Rick & Morty's Smart Robot

Today's Machine Learning systems belong to the group of ' weak AIs'. As we have seen, these Intelligences are only capable of understanding patterns and making simple deductions. They are very useful to support us in certain contexts, but they are not autonomous systems at all.

On the other hand we would have the 'strong AIs' , those represented in futuristic stories where they are equal to or much more intelligent than humans. We can find notable examples in popular culture like 'Matrix' , 'Terminator' , 'Ghost in the Shell' or 'Halo' . In fact, in this list there are two works that are related to each other; Guess which ones?

Today we are still developing fully autonomous and safe cars . We are continuously advancing, but we still have a way to develop an equal fact made entirely of technology.

If you want to know more about it, you can visit our article about Artificial Intelligence . It is a text from a more general point of view and we study a bit the possible ramifications that this technology will have.

Final Words on Machine Learning

Similar to our conclusion on Artificial Intelligence, it is clear that the future is uncertain. However, it is inevitable that evolution will need to be reviewed to implement technology among its skills and characteristics.

Little by little, the Internet will be more and better controlled by programs and algorithms. Social networks will be better calibrated and will offer us content more according to our tastes. And finally, online relationships will be much more secure by detecting more easily when there is a danger of fraud or the like.

On the other hand, do not be surprised that this century is when the IoT (Internet of Things) will shine. It is an idea that we have been dreaming of for a long time and that is getting closer. In addition, the IoT is a large bidder of cutting-edge technologies related to Machine Learning, although it still lacks some adjustments regarding security.

For our part, we think it will be a gradual evolution and as long as you are informed of what is happening, you have nothing to fear. New cars or refrigerators may sound strange to you, but I certainly don't think we'll see the awakening of 'strong AIs'.

We recommend reading the best laptops on the market

Finally, we have to confess that we are not experts in Artificial Intelligence or Machine Learning , so do not be surprised by some strange data. If we have made a mistake, don't hesitate to tell us! After all, we are not yet perfect machines.

And you, what do you think of Machine Learning and Artificial Intelligence ? In which aspect do you think they should be implemented? Share your ideas below.

Clever Dataapdsaslagacetawhatsnew font