Each of us experiences contact with artificial intelligence every day. We reach for its possibilities spontaneously and often without realizing that it is to artificial intelligence that we owe the comfort of life and a number of conveniences. Who doesn’t use Gmail, Google Translate, a smart personal assistant or Netflix recommendations? We also watch dozens of ads every day, which are often selected for us by algorithms. How has the use of artificial intelligence evolved over the years?
In this article you will learn:
- How applications of artificial intelligence were initially used
- How businesses use AI today
- Several known practical applications of AI
- What using AI could bring in the near future
Table of Contents:
- Initial motives for using AI
- The evolution of AI on the example of chess duels
- Where are artificial intelligence algorithms used today?
- Examples of AI applications
- What awaits AI in the future?
Initial motives for using AI
Although it may seem that artificial intelligence has entered our lives only in recent years, the first significant work in this field saw the light of day as early as the middle of the twentieth century. British mathematician Alan Mathison Turing, the creator of the so-called Turing machine, a device used to perform algorithms, is considered to be the father of artificial intelligence. Shortly thereafter, the first chess or checkers programs and others capable of intelligent behavior in specific situations came into existence.
The early 1980s saw the separation of artificial intelligence and machine learning, which until then had served as a training program for artificial intelligence. Artificial intelligence research thus moved away from algorithms and began to focus on using logical knowledge-based approaches. Machine learning, on the other hand, has become a separate field that focuses on solving practical problems. The dynamic development of this area can be attributed to not only a large number of researchers and neural network specialists, but also the growing popularity of the Internet, which not only allowed its solutions to be shared with a wider audience, but also significantly facilitated access to digital data.
Scientists did not have to wait long for the surprising results of their work. Soon, the first autonomous vehicles appeared, programs able to draw like a human, translate into many languages or process natural language. In time, the first solutions allowing for speech or face recognition appeared. In the following years their precision and capabilities were significantly improved, thanks to which today artificial intelligence is widely used in almost every aspect of our lives.
The evolution of AI on the example of chess duels
The beginnings of artificial intelligence development were simple programs that were adapted to the capabilities of computers at the time. It is not surprising that one of the first programs created by the pioneers of artificial intelligence was a program to play chess. This simple and extremely popular game has become a kind of witness to the evolution of artificial intelligence. It is therefore an excellent example to better understand the differences between artificial intelligence (AI), machine learning (ML) and deep learning (DL).
In the case of artificial intelligence (AI), it is possible to write a program which will, on the basis of some predefined logic rules, play chess. But the program doesn’t necessarily have to take into account any additional data like the knowledge gathered from previous game sets, etc.
In the case of using machine learning (ML), the algorithm playing chess will, apart from the predefined rules (as it is in general AI example), learn from different master players’ moves, check parts of the game, etc. In most cases the overall ‘chess game problem’ will be divided into subproblems, like:
- how to optimally begin the game (probably based on predefined rules – chess openings),
- how to optimally solve middle-game problems (usually based on some data like different grandmasters plays to evaluate specific positions),
- how to optimally solve end-game problems (when there is a small number of pieces on the board, there is usually a possibility for exact simulation of each move till the end, to choose the optimal next move).
So there is usually a mix of different techniques optimized towards solving some subproblems, in order to solve a higher-level problem. Specific machine learning algorithms can be used on some level. In that case the more data from previous sets it has, the more and faster it can learn how to win the game.
Deep learning (DL) algorithms allow not only to solve specific, simpler subproblems but can also learn what the dependencies between those subproblems (layers) are, so those algorithms are capable of solving the whole higher-level problem without manually dividing it and solving each subproblem independently. In the “chess game example” such an algorithm could ideally learn only from a set of historical chess games and could automatically detect that there is a need to divide those plays into parts, like the beginning, mid- and end-game. The ideal algorithm would learn how to optimally play the end-game. Based on that it would solve the middle-game problem and based on that it would also solve the problem of how to begin the game in order to have the best mid-game position. And all that without human justification, manually implemented additional rules and so on.
Where are artificial intelligence algorithms used today?
Today, artificial intelligence accompanies us almost at every step. It is used not only for the development of autonomous vehicles and robots, but also to conquer space, explore the galaxy, support medics in identifying disease entities and disturbing changes in X-rays, as well as solving many problems faced by business every day.
We are talking about processing huge data sets and supporting management in making key decisions, recommending optimal solutions, forecasting supply and demand, dynamic pricing or inventory optimization. This also applies to broadly understood marketing activities. Today, artificial intelligence decides not only what banner and content to present to the user, but also analyzes all the available information to decide whether it makes sense and will result in the desired action. AI also stands behind the best recommendation engines that tells us every day what songs, movies or products we may like.
Artificial intelligence allows us not only to improve, but also automate many business processes and thus generate tangible benefits for the organization. No one should be surprised, therefore, that business is so eager to reach for new technology and find more and more new possibilities of its application.
Examples of AI applications
How often artificial intelligence enters our lives is perfectly illustrated by specific examples of its use. We have probably already had the opportunity to come across some of them, but it is very possible that we did not even realize that the specific solutions or improvements are due to artificial intelligence. What specific solutions are we talking about?
In May 2018, CEO of Google presented its newest feature – Google Duplex. The AI assistant that is able to make normal phone calls, make arrangements and check information about businesses (e.g. business hours or the status of in-demand inventory). In one demo, the Google assistant goes back-and-forth with a restaurant’s clerk, with the machine system parsing the verbal exchange to book an available seat. Due to the regulations in force, it took some time to implement Google Duplex across the USA, but today the service based on artificial intelligence is available in 49 US states. Google is gradually making the service available also in other countries, including Canada, UK, Japan, Spain, France and others.
LG’s ThinQ and Deep ThinQ AI
LG ThinQ products and services have the ability to employ deep learning and communicate with one another, utilizing a variety of AI technologies from other partners as well as LG’s own AI technology, DeepThinQ. LG’s own AI platform, DeepThinQ, is designed to give products brains, so they can predict user needs and adapt different behaviors accordingly. The fridge with DeepThinQ tech can analyze usage and eating patterns. It will only make ice at the times of day it is needed or change sterilization system patterns when it detects higher temperatures externally.
This solution can be connected with e-commerce stores for making shopping easier and more automated.
Gaming industry – deep learning to improve graphics
Expert Chintan Trivedi tried to find out whether the recent developments in deep learning can help to improve the graphics while simultaneously also reducing the efforts required to create them. He tried to focus on improving the players’ faces in FIFA using the (in?)famous deepfakes algorithm. This is a Deep Neural Network that can be trained to learn and generate extremely realistic human faces. His focus in this project lies on recreating the players’ faces from within the game and improving them to make them look exactly like the actual players.
What is the result of the work of AI experts on FIFA? HyperMotion is already up and working, but only on the PlayStation 5 and Xbox Series S/X versions of FIFA 22. It brings more fluidity and realism while playing and at the same time more accurately simulating real-life football.
This solution, after a lot of development, focusing on recreating actual bodies (not only faces) could be used in the fashion sector, while making visualizations for trying on clothes. This is a nice and very visual example of how deep learning can improve something we already thought was almost excellent.
Gaming industry – deep learning to improve quality
Electronic Arts, the company behind FIFA, among others, also uses artificial intelligence to test its games. By introducing a self-learning mechanism into game testing structures, player behavior can be replicated and many annoying bugs identified. As a result, experienced testers can focus on fundamental testing and leave the most tedious and time-consuming tasks to technology.
„With neural networks learning how to play the games by themselves we can mimic human-like behaviors that go beyond current automated testing approaches, of scripted bots, where the game learns and plays like a human. This means we find bugs faster, in ways that humans do, before they get out in the wild to millions of players” – Colin Barre-Brisebois, EA SEED Head of Technology.
You can test your store’s performance in the same way and quickly identify issues that can have a very negative impact on conversions or revenue.
Initially, Google’s linguistic data was obtained from the United Nations and European Parliament documents. At present, AI stands behind the high and still growing accuracy of translations. The huge step was taken in 2016 when Google implemented deep learning methodology – Neural Machine Translation (NMT) which results in using a wide variety of linguistic sources. High quality training data taken from the Internet (after reducing the noise in data) allows Google to constantly improve models that are in use and reduce more and more errors in translation.
From the e-commerce perspective, this kind of a solution can be a real game-changer in terms of opening e-commerce services into different markets. For instance if some European e-commerce would like to open to the Japanese market, it would need to translate not only static elements of the website (which is quite easy even without a good automatic solution), but it would also have to maintain the translations of thousands of rotating offer names, their descriptions and users reviews, which would practically be effortless if great translation solutions were available. In the future, deep learning algorithms could become even better than human language translators, similarly to how they became better than humans in handwriting recognition problems.
What awaits AI in the future?
Artificial intelligence is already revolutionizing our daily lives and the way many industries and businesses operate. Moreover, there is no indication that this trend will reverse in the near future. On the contrary, artificial intelligence has become a catalyst for the development of many new technologies (including robotics, big data and IoT) and everything indicates that this role will continue in the coming years.
And although today certain areas and aspects of our lives still seem to be virgin territory for the development of artificial intelligence, we should expect that this may soon change. Of course, this does not mean that those areas that already boast numerous examples of the application of artificial intelligence will stop witnessing the dynamic development of the technology. Scientists will continue to put a lot of effort and energy into developing autonomous vehicles, improving robots that can replace humans in many repetitive tasks in the manufacturing industry, or more accurately diagnosing disease entities.
The marketing industry is also likely to face major changes – we can expect even better-suited ads to the context of the search, more relevant recommendations and personalization taken to a completely new level, which from the business perspective will result in better customer experience and increased sales.
If you have any questions, comments or issues, or you’re interested in meeting with us, please get in touch.