Recently we hear more and more about artificial intelligence. It is used almost everywhere: from high technology to complex mathematical calculations to medicine, the automotive industry and even the use of smartphones. The technologies underlying KI's work in the modern sense are used every day and sometimes we may not even think about it. But what is artificial intelligence? How does he work And is it dangerous?
1 What is artificial intelligence?
2 How Artificial Intelligence Works
3 deep learning and neural networks
4 Neural networks – is it an artificial human brain?
5 Why deep learning and neural networks are used
6 The Limitations of Deep Learning and Neural Networks
7 The Future of Deep Learning, Neural Networks and AI
What is artificial intelligence?
Let's first define the terminology. If you think of artificial intelligence as something that can think independently, make choices, and generally display awareness, then we hurry to disappoint you. Almost all existing systems today do not even "come close" to such a definition of AI. And those systems that show signs of such activity actually all work within pre-defined algorithms.
Sometimes these algorithms are very, very advanced, but they remain the "framework" in which the AI works. There are no "freedoms", and even more, machines have no signs of consciousness. These are just very productive programs. But they are "the best in their field". In addition, the AI systems continue to improve. Yes, and they are arranged quite unanal. Although we reject the fact that modern AI is anything but perfect, it has much in common with us.
How does artificial intelligence work?
First and foremost, the AI can do its job (about it later) and gain new skills thanks to in-depth machine learning. We often hear and use this term. But what does he mean by that? Unlike the "classical" methods, machine learning algorithms let the system develop independently and study the information available when all required information has been preloaded into the system. Whereby the device can search independently in some cases.
For example, to create a fraud detection program, the machine learning algorithm works with a list of banking transactions and their final outcome (legal or illegal). The machine learning model examines examples and develops a statistical relationship between legitimate and fraudulent transactions. If you subsequently provide the data of a new banking transaction to the algorithm, it classifies it using the templates that it outlined in advance using the examples.
The more data you provide, the more accurate the machine learning algorithm will be in performing its tasks. Machine learning is particularly useful in solving problems where the rules are not pre-defined and can not be interpreted in a binary system. Back to our banking example: In fact, we have a binary calculation system at the output: 0 – legal business, 1 – illegal. However, to come to this conclusion, the system has to analyze a whole series of parameters. If you enter them manually, it takes more than a year. And predicting all options will not work anyway. And a system based on deep machine learning will be able to detect something, even if it has not been hit exactly in such a case before.
Deep Learning and Neural Networks
While traditional machine learning algorithms solve many problems in which there is much information in the form of databases, they have not coped well with so-called "visual and audio" data such as images, videos, sound files, and so forth.
For example, creating a model for predicting breast cancer using classical machine learning methods requires the efforts of dozens of medical experts, programmers, and mathematicians, "said AI researcher Jeremy Howard. The scientists would have to develop many smaller algorithms so that machine learning can handle the flow of information. A separate subsystem for x-ray examination, a separate one for MRI, another for the interpretation of blood tests and so on. For every kind of analysis we need our own system. Then they would all be grouped together into one big system … This is a very difficult and resource intensive process.
Deep learning algorithms solve the same problem using deep neural networks, a kind of software architecture that is inspired by the human brain (although neuronal networks differ from biological neurons, the working principle is almost the same). Computer neuronal networks are communications of "electronic neurons" capable of processing and classifying information. They are a kind of "levels" and each "level" is responsible for something own and ultimately forms the overall picture. For example, when you train a neural network on images of various objects, you can find ways to extract objects from those images. Each layer of the neural network discovers certain features: the shape of objects, colors, the appearance of objects, and so on.
The surface layers of neural networks have common features. Deeper layers already reveal actual objects. The diagram shows a simple neural network. Green means input neurons (incoming information), blue means hidden neurons (data analysis), yellow means output neurons (solution)
Are neural networks an artificial human brain?
Despite the similar structure of the machine and human neural networks, they have no signs of our central nervous system. Computer neuronal networks are essentially the same utilities. By chance, our brain proved to be the best organized system for performing calculations. You probably heard the phrase "our brain is a computer"? Scientists simply "reiterated" some aspects of their structure in a "digital form." This only allowed to speed up the calculations, but not to make the machines conscious.
That's interesting: when does artificial intelligence learn to argue?
Neural networks exist since the 1950s (at least in the form of concepts). Until recently, however, they did not receive much development work because their creation required enormous amounts of data and processing power. In recent years, all of this has become available, so that the neural networks have come to the fore after they have received their development. It is important to understand that there was not enough technology for their complete appearance. How they are now not enough to take technology to a new level.
What are Deep Learning and Neural Networks used for?
There are several areas where these two technologies have contributed to significant progress. We also use some of them daily in our lives and do not even think about what's behind them.
Computer vision is the ability of software to understand the content of images and videos. This is an area where deep learning has made great progress. For example, deep learning image processing algorithms can detect various types of cancer, lung disease, heart disease, etc. And faster and more efficient than doctors. However, deep learning is also rooted in many of the applications you use every day. Apple Face ID and Google Photos use Deep Learning to recognize faces and improve image quality. Facebook uses Deep Learning to automatically tag people in uploaded photos, and so on. Computer Vision also helps organizations automatically identify and block questionable content such as violence and nakedness. And finally, deep learning plays a very important role in enabling self-driving cars to understand what surrounds them. Speech and speech recognition. When you issue a Google Assistant command, deep learning algorithms convert your voice to text commands. Some online applications use Deep Learning to transcribe audio and video files. Even if you "shazamit" a song, neural network algorithms and deep machine learning play a role. Searching the Internet: Even if you're looking for something in a search engine, companies have begun to connect neural network algorithms to their search engines to make your query clearer and search results as accurate as possible. The power of the Google search engine has increased many times over, after the system has been switched to deep machine learning and neural networks.
The Limits of Deep Learning and Neural Networks
Despite all advantages, deep learning and neural networks also have some disadvantages.
Data dependency: In general, deep learning algorithms require a large amount of training data to perform their tasks accurately. To solve many problems, unfortunately, insufficient training data is required to create working models. Unpredictability: Neural networks are developing in a strange way. Sometimes everything goes as planned. And sometimes (even if the neural network does its job well) even the developers have difficulty understanding how the algorithms work. The lack of predictability makes it extremely difficult to eliminate and correct errors in the neural network algorithms. Algorithmic Bias: Deep learning algorithms are as good as the data on which they are trained. The problem is that the training data often contains hidden or explicit errors or omissions and the algorithms "inherit" them. For example, a face recognition algorithm that is primarily trained on white person photographs will be less accurate for people of a different skin color. Lack of generalization: Deep learning algorithms are well-suited for performing specific tasks, but they do not generalize their knowledge adequately. Unlike humans, the deep learning model that was trained to play StarCraft can not play any other similar game, such as in WarCraft. In addition, deep learning is not suitable for processing data that differs from the training examples.
The future of deep learning, neural networks and AI
It is clear that work on deep learning and neural networks is far from over. Various efforts are being made to improve deep learning algorithms. Deep learning is an advanced method of creating artificial intelligence. It has become increasingly popular in recent years due to the wealth of data and increased computing power. This is the core technology behind many of the applications we use every day.
But will consciousness ever be born on the basis of this technology? Real artificial life? Some scientists believe that the moment the number of connections between the components of artificial neural networks approaches the same indicator that exists in the human brain between our neurons, something can happen. This statement is very doubtful. For a true AI to appear, we need to rethink the approach to creating systems based on AI. Now there are only application programs for a narrow range of tasks. How could we not believe that the future has already come …
What do you think people will create an AI? Share your opinion in our telegram chat.