What is artificial intelligence?
What is artificial intelligence?
Artificial intelligence (AI) is an extensive industry of computer sciences, which is focused on creating smart machines that can perform intellectual tasks.
To date, there are many approaches to creating algorithms. Achievements in areas of machine and deep learning over the past few years have significantly changed the technological industry.
What are the definitions of artificial intelligence?
The fundamental goal and vision of artificial intelligence managed to establish English mathematics Alan Turing in the article “Computing machines and mind”, published in 1950. He asked a simple question: “Can cars think?"". Then the scientist proposed the famous test named after him.
At its core, it is a branch of computer sciences that seeks to answer the question of Turing in the affirmative. This is an attempt to reproduce or simulate human intelligence in machines.
The global goal of artificial intelligence still raises many questions and disputes. The main restriction in the definition of AI as simply “reasonable machines” is that neither scientists nor philosophers can explain what artificial intelligence is and what exactly makes the machine smart.
Scientists and authors of the textbook “Artificial intelligence: modern approach” Stuart Russell and Peter Norvig combined their work around the theme of intellectual agents in cars and defined AI as “the study of agents who receive perception from the environment and perform actions”.
During the performance at Japan Ai Experience in 2017, Datarobot Director General Jeremy Achin began his speech with the following definition of how artificial intelligence is used today:
“AI is a computer system that can perform tasks that require human intelligence … Many of these systems work on the basis of machine learning, others on the basis of deep training, and some of them are on very boring things, such as rules”.
Although these definitions may seem abstract, they help to determine the main areas of theoretical research in the field of computer sciences and provide specific ways to implement programs with artificial intelligence to solve applied problems.
What is the merit of turing before the development of AI?
In the middle of the last century, Alan Turing laid out a theoretical base that was ahead of its time and formed the basis of modern computer sciences, for which he was called the "Father of Informatics".
In 1936, the scientist created an abstract calculator – the so -called Turing machine – an important component of the theory of algorithms, which formed the basis of modern computers. In theory, such a machine can solve any algorithmic problem.
In turn, if the algorithm can be launched by a Turing machine, then the programming language used to create it will have “fullness by turing”, on which you can write any algorithm. For example, the C# language has this fullness, and HTML does not.
Also, a mental test is named after mathematics, which is not related to the machine, but is directly related to artificial intelligence – Turing test. In the scientific environment, it is believed that as soon as the car passes this test, it will be possible to fully talk about the appearance of reasonable machines.
The essence of the game is that a person with the help of text correspondence interacts simultaneously with the machine and other person. The task of the computer is to mislead the participant in the test and convincingly impersonate a person.
What happens?
Artificial intelligence is usually divided into two large categories:
- Weak AI [Weak AI]: This type of artificial intelligence, which is sometimes called "narrow AI" [narrow ai], works in a limited context and is an imitation of human intelligence. Weak AI is often focused on a very good performance of one task. And although these machines may seem smart, they work with great restrictions.
- General artificial intelligence (Agi): AGI, sometimes called “strong AI” [Strong ai], is a form of artificial intelligence that we see in films, for example, robots from the “Wild West World” or Joy’s hologram from “Running Blade 2049”. Agi is a machine with general intelligence, which, like a person, can use it to solve any problem.
What is weak artificial intelligence?
Weak AI surrounds us everywhere, and today it is the most successful implementation of artificial intelligence.
Focusing on the implementation of specific tasks, over the past decade, he has made many breakthroughs that brought “significant social benefits and contributed to the economic viability of the nation”, according to the report “Preparation for the future of artificial intelligence” published by the Obama administration in 2016.
Here are a few examples of weak AI:
- Search for Google;
- Software for recognition of images;
- Siri, Alexa and other voice assistants;
- Unmanned cars;
- Netflix and Spotify recommendation systems;
- IBM Watson.
How weak AI works?
Most of the weak AI is based on achievements in the field of machine learning and deep learning. The similarity of these concepts can be confused, but they should be distinguished. Venture capitalist Frank Chen proposed the following definition:
“Artificial intelligence is a set of algorithms that are trying to imitate human intelligence. Machine training is one of them, and deep training is one of the methods of machine learning. ".
In other words, machine learning supplies the computer with data and uses statistical methods to help it learn to perform tasks without being specially programmed for them, which eliminates the need for millions of lines of the written code. Popular types of machine learning are training with a teacher (using marked data sets), training without a teacher (using nemarkized data sets), and training with reinforcement.
Deep training is a type of machine learning in which the entered data is processed through the architecture of a neural network based on biological principles.
Neural networks contain a number of hidden layers through which data is processed, which allows the machine to “delve” into their training, establish connections and weigh the input to achieve the best results.
What is machine learning?
Artificial intelligence and machine learning are not the same. Machine training is only one of the subsections of AI.
The most common types of machine learning – with a teacher, without a teacher and with reinforcement.
Training with the teacher
Use when the developers have a marked data set and they know what signs should look for the algorithm.
As a rule, it is divided into two categories: classification and regression.
The classification is used in cases where it is necessary to attribute objects to pre -known classes. This type of learning is used in spam filters, language determination, or identifying suspicious transactions.
Regression is used when it is necessary to correlate an object with a temporary line, for example, to predict the value of securities, demand for goods or making medical diagnoses.
Training without a teacher
A less popular species due to its unpredictability. Algorithms are trained on unreasonable data and they need to independently find signs and patterns. Often used for clustering, reducing the dimension and searching for associations.
Clastorization is like a classification, but without well -known classes. The algorithm must itself find signs of similarities in objects and combine them into clusters. Used to analyze and mark new data, compression of images or combining the marks on the map.
Reduction of dimension – summarizes specific features in abstraction of a higher level. It is often used to determine the topics of texts or create advisory systems.
Associations have found their application in marketing, for example – in the preparation of promotions and sales or analysis of user behavior on the site. Can also serve to create a recommendatory system.
Reinforcement training
This is an agent’s training to http://coin-graph.website/?p=368 survive in an environment in which it exists. Wednesday can be anything: from video games to the real world.
For example, there are algorithms that play super-Mario no worse than people, and in the real world autopilot in Tesla or robotesol cars do everything to go around obstacles in their path.
Reinforcement training provides for an agent for the correct action and punishment for errors. The algorithm does not have to memorize all its previous experience and calculate all possible options for the development of events. He must learn to act on the situation.
Remember when the car beat a person in th? Long before that, scientists have found that more variations in this game are more than atoms in the universe. No computer program from now existing could calculate all the options for the development of the party. However, Alphago, the Algorithm of Google, coped with this task without calculating all the moves in advance, but acting in circumstances, doing this with incredibly high accuracy.
What are neural networks and deep training?
The concept of artificial neural networks is not new. For the first time, this concept was formulated by American scientists Warren McCalllock and Walter Pitts in 1943.
Any neural network consists of neurons and connections between them. Neuron is a function that has many inputs and one exit. They exchange information among themselves through communication channels, each of which has a certain weight.
Weight is a parameter that determines the strength of the relationship between neurons. Neuron himself does not understand what he sends, so the weight is necessary in order to regulate which entrances to respond, and which does not.
For example, if the neuron sends the number 50, and indicate the weight of the communication 0.1, then the result will be 5.
As the architecture of neural networks complicates, the neurons decided not to bind not as they like, but by layers. Within the layer, neurons do not interact in any way, but receive and transmit information from the previous layer to the next.
As a rule, the more layers in the neural network – the more complicated and more precisely the model. But then, 50 years ago, the researchers rested on the limitations of computing power. As a result, the technology turned out to be disappointed and forgot about it for many years.
They remembered her in 2012 – students of the University of Toronto Alex Krizhevsky, Ilya Satskever and Jeff Hinton won the Imagenet computer vision competition. They used a spanning neural network to classify images, the error level of which was 15.3%, which is more than 10% lower than that of the team that took second place. The revolution in the field of deep learning occurred largely due to the development of graphic maps.
Deep learning differs from neural networks only in large -sized network training methods. In practice, as a rule, the developers do not find out which network can be considered deep and which. Today, even to build five layers, developers use “deep” libraries such as Keras, Tensorflow or Pytorch.
To date, the most popular networks are sparkle neural networks (CNN) and recurrent neural networks (RNN).
CNN is often used to recognize faces, search for objects in photographs and videos, improve the quality of images and other tasks. Recurrent networks are found in machine translation of text and speech synthesis. For example, since 2016, Google Translate has been operating on the basis of RNN architecture.
Generative and consistent networks (GAN) also found popularity. It is based on two neural networks, one of which generates data, for example, an image, and the second is trying to distinguish the correct samples from the wrong. Since two networks compete with each other, an antagonistic game arises between them .
GAN is often used to create photo realistic photos. For example, the repository of the images of This Person Does Not Exist consists of portrait photos of “people” created by a generative neural network.
What is general artificial intelligence?
The creation of a machine with an intelligence of the human level, which can be applied to any task, is a holy grave for many AI researchers, but the search for Agi is associated with some difficulties.
General AI has long been a muse of anti -utopian science fiction, in which supernatural robots flood humanity, but experts agree that this is not what we need to worry in the near future.
American inventor and futurologist Ray Kurzweil predicted that the total AI will appear by 2029. His colleague Rodney Brooks is not so optimistic, and is sure that the turning point in the development of the technology of the machine mind will occur by 2300.
Stuart Russell, one of the authors of the textbook “Artificial Intelligence: Modern approach”, suggests that the invention of Agi will become accidental, such as the discovery of nuclear energy in 1933. The scientist believes that this is a vivid example of how it is pointless to give any forecasts in the development of such an unpredictable technology that has not yet been fully studied.
Subscribe to FORKLOG news at Telegram: Forklog AI – all news from the world of AI!
