▷ Artificial intelligence: what is it and current practical examples?

For a few years, companies have continually spoken to us about Artificial Intelligence that they introduce in their services, applications and processors. However, although they bear the same name, thank God, the Artificial Intelligence of our washing machine (for reasons that escape us) and of our smartphone is not so developed as to make them reflect on their existence and our power over them. For now…

As we already told you in the article about the AI ​​development USB Intel Movidius, Artificial Intelligence is here to stay and help us solve day-to-day problems. But what exactly is Artificial Intelligence?

Source: Source Dexeter

The gif you see above shows in a very simplified way how a deep neural network works. These systems require hard training to later be able to, for example, recognize images, optimize solutions or simply learn more. In essence it is a set of algorithms that we could categorize as AIs and that belong to the field of Deep Learning.

Index of contents

Artificial Intelligence: new programming

Today, Artificial Intelligence does not make up intricate mixed systems of technology with a conscience as is often seen in science fiction works. The one we create falls rather on the definition of complex algorithms that return results based on the inputs and the commands that have been taught to them. Although that's just one of the meanings there is.

There are different ways of understanding Artificial Intelligence, but we could divide it into four main groups:

AIs that think like humans

Butter robot Rick and Morty

Complex computer systems with their own conscience that think and decide according to their desire and exceed the characteristics for which they were programmed ( Ghost in the Shell). It is not yet within our reach and we do not even know if it will be possible in the future, so there is not much to comment.

IAs that act like humans

Thinking like a human is not the same as pretending to act like a human. Today we create some systems like these where randomness and concrete functions are introduced to give the feeling that Intelligence thinks like a person.

Pepper smart assistant

In video games we see this continuously, as machine-controlled enemies often seek to simulate human-like behaviors. Separated from video games, it has been achieved that an Artificial Intelligence can write with imperfections and irregularities as a person would.

IAs that think rationally

Possibly the most common version of this technology today. We say that they think rationally because we give them the tools to offer efficient and meaningful results. They are able to adapt to the environment they are in easily, although they are far from thinking for themselves.

AlphaStar Learning

An example of this is Artificial Intelligence that plays video games like AlphaStar (StarCraft II) or AlphaZero (chess, shogi and go). These machines are even capable of fighting human opponents and have already defeated the occasional world champion.

IAs that act rationally

Since they 'act' we discover that they do not process the data we pass on to them, they only appear to think rationally. This is the most simplistic version of this technology and it is a stage that we have already largely passed. Some computer systems resort to this technology, since it is much easier to program and their work is usually simple.

Smart Vacuum Cleaner

For example, the machines that receive calls and guide you through their options or the intelligent assistants of web pages, which usually ask you to recommend related solutions.

Already having an acceptable image of how the Intelligence is distributed according to how complex they are, let's take you to the heart of the matter.

The mathematics of thought

One of the ways to program Artificial Intelligence is handling the data as imaginary units called tensors. Tensors are a complex algebraic unit (of scalars, vectors, and matrices) that requires knowledge of mathematics to work properly with them. Consequently, the performance of AI applications will be as good as mathematical manipulations of the data have been performed.

Simplified explanation of turnbuckles

To expand the development of this type of software, many groups have created and opened their code libraries to the public to cooperate and create, together with the community, more intelligent systems. TensorFlow by Google, CNTK by Microsoft, Theano, Caffe2 and Keras are some of the most relevant examples. Each of the libraries focuses on the problem from different angles and thanks to this we have at our disposal the development of AI at different levels of abstraction.

If you don't know what levels of abstraction are, it is a system that measures how closely a computer language is to spoken language. The higher a level of abstraction, the more it resembles a human language and the lower, the more machine code, that is, that world that works only with zeros and ones.

New systems, new hardware

It is clear that all software runs within hardware, however, it is easy to fall into the illusion that the cloud can cope with everything, but the reality is not so sweet. Depending on how the code is optimized, it may be the case that the AI ​​works locally (on the smartphone, PC or Internet of Things device). Or the devices can be allowed to send the calculations to the servers, process it, and these return the result.

Cloud services

In many cases , the “small” device tries to carry out a large part of the calculations locally and sends only part of the problem to the server, thus saving many service management costs.

Artificial Intelligence in the day to day

We know that thinking about the future of this is something very interesting and for some even exciting, but you don't have to go that far to see the first fruits. Where can we find traces of Artificial Intelligence in today's society?

Artificial Intelligence on mobile

It may seem to go unnoticed, but it surrounds us on all sides. Starting with home devices, new mobiles often have small built-in systems called Artificial Intelligence that help you take better photos. Selectively focus, post-process images to make them look sharper, more colorful, or contrasty. Some are even able to recognize the objects we capture and offer us related searches.

In this field , the colleague who is 'OK Google' away, who learns from everything we tell her and is capable of processing infinite requests, also stands out. While we can find you 'machined' very easily (like not being able to carry on a conversation), we can't dismiss the hard work we know is behind it.

Google Assistant

We also have to talk about the imminent autonomous driving. Cars like Tesla already offer those AI-controlled alternatives in some countries. These systems are capable of capturing the environment around the car, processing prohibitions, hazards, and so on, and driving safely accordingly.

Although we do not need to go to such high levels of intelligence in the automotive world. We can see that some cars already have such interesting systems as emergency stop detection or automatic parking.

The queen in the shadows:

By now you may already be thinking that AI is everywhere and that at any moment they are rebelling, but rest assured, your toaster is not going to kill you while you sleep. What we can confirm is that this technology controls more than you think and is responsible for many of the trends in society.

Youtube, Twitter, Google ads… All this is controlled to a certain extent by the settings you have indicated, but also by Artificial Intelligence that decide what to show you. Do you hear a message similar to: "I want to share my data with Google so that it offers me ads that may interest me" ?

How does this work? Well, you will see, based on what you consume on the Internet, a profile is created with your tastes and you are related to many other people. When Internet services need to show you something, they use this profile made up of millions of individuals to estimate what may interest you.

Simplified Big Data explanation

This way of analyzing huge amounts of data (Big Data) using AIs is taking a lot of strength and careers are appearing all over the world ready to prepare the future on this subject. As you will understand, the data that users use is counted by TeraBytes every second, so a person is not able to analyze them all. This is where Artificial Intelligence works with the data and it is people who use it to make estimates and so on, using, for example, statistics.

WE RECOMMEND YOU Google Home Mini: what it is and what it is for, features

The Foundation: Deep, Machine Learning

We are going to navigate a bit in the video game world to understand Deep Learning a little better, since AI has entered the field of video games both as a player (as we mentioned before), and as a programmer and designer. If you follow the advances of the industry, NVIDIA has been gaining notoriety for different technologies among which is its DLSS (Deep Learning Super Sampling) system, an Artificial Intelligence that is capable of rescaling images.

DLSS comparison

The function of DLSS is to transform an image from FullHD (1080p) to UltraHD (4k) to be able to play the most demanding titles with better frame rates. At first, users complained that the images looked blurry and out of focus, but just a few months later the results are great.

This is thanks to Deep Learning, a system through which Artificial Intelligence learns with practice and error. In the case of DLSS, NVIDIA Intelligence was continuously analyzing images in UltraHD resolution and trying to recreate them using a FullHD image as a basis. In other words, it is as if they gave you a quarter of an image and you had to fill in the gaps that you don't know. Deep Learning is a type of system belonging to what is called Machine Learning or Automatic Learning in Spanish.

Machine Learning and Deep Learning

Machine Learning could be classified as the foundation stone of Artificial Intelligence. These are different sets of algorithms that are often used for machines to learn tasks, among other things. For example, recognizing an image, playing chess or detecting moods are challenges that can be learned and different types of algorithms are used depending on the challenge.

Machine Learning is said to be the set of algorithms that allow a machine to learn from the experience it is accumulating. On the other hand, Deep Learning focuses on learning with heterogeneous inputs. Both disciplines are being developed and studied with energy since the future of Artificial Intelligence is uncertain.

The future of Artificial Intelligence

From our perspective, the possibilities of Artificial Intelligence seem endless. We still don't know what our limit is and we are already working on creating another being similar to us, but what can we expect in the future?

Nothing that we will comment can be taken for granted, but they are statements based on certain arguments derived mainly from observing how these machines have evolved.

Internet

First of all, it seems inescapable that we are moving towards a world dominated by the Internet, which is why AIs will increasingly have more relevance and power over the medium. It is not something that should scare us, since this is the only way in which we could ensure the maintenance of the platform. With this we could surf the web in a somewhat more guarded space, but at the same time much safer. As first pioneers of this we have Facebook bots that analyze and estimate if suicidal thoughts run through you and if they detect it, they contact you.

Likewise, in the physical world, autonomous and assisted cars will become increasingly dominant until the moment when driving is only recreational. Perhaps the change does not happen for a hundred years, but the change will happen.

Another change that is also predicted is the exchange of hard work for machines. It is a revolution that many fear, but it seems inevitable, so we will have to be prepared.

Cyborg Neil Harbisson

And although it seems something typical of science fiction, it is very likely that in the future we will have to find ways to integrate technology and Artificial Intelligence in our body. In fact, the first cyborg in history already exists and is called Neil Harbisson.

Beyond this shore the sea of ​​ideas is immense. Who knows? Perhaps the machines of a factory all work in unison under the command of a chief machine with the primitive machine-machine languages. Perhaps one day the best stock market speculator will be an Artificial Intelligence or even the best motoGP biker.

Artificial intelligence

It may seem like a strange, scary future, but we certainly have other problems to try to solve!

And what do you know about AIs? Are you eager to see what will come? Tell us what your ideas are about Artificial Intelligence.

PowerDataIberdrolaIndraNewsRoom Font