Through AlphaGo, artificial intelligence attracted attention from around the world. After a long period of stagnation, artificial intelligence entered a turning point with the rapid development of cloud computing environments and a dramatic breakthrough in implementing deep learning supported by big data. Now artificial intelligence has emerged as a key element of the 4th Industrial Revolution.
Resurrection of artificial intelligence
Perhaps the list of important events representing 2016 will surely include the match between AlphaGo and Lee Sedol and artificial intelligence. Through AlphaGo, artificial intelligence has emerged as a field of interest to the entire nation. Thanks to this, education and promotion of artificial intelligence were accomplished at once.
The idea of creating thinking machines with human intelligence has a long history. Artificial intelligence emerged naturally as humans developed computation methods. If you look at the history of artificial intelligence, it has gone through a considerable period of ups and downs. Beginning in the 1930s and 1940s, expectations for thinking machines began to grow in earnest. The Turing Test was developed, and theories about human thought processes such as mathematical logic, computation, cybernetics, and information theory emerged.
The term artificial intelligence (AI) first appeared in 1956. However, creating a general intelligence program turned out to be more difficult than expected, leading to a recession until the 1970s. The UK’s AI research institute was disbanded, and US research foundations stopped supporting AI research.
Then, in the 1980s, artificial intelligence was rediscovered through neural net theory. Neural network theory is a theory that views human thinking as a product of brain activity and starts from the theory that a thinking machine can be created by analyzing this brain structure and identifying the processing mechanism. However, there was no way to manage the vast amount of data to apply this theory. It hit a recession again, but was revived again in the 1990s with the development of the Internet.
This is because it has become possible to collect vast amounts of data through search engines, etc. Here, machine learning has become possible, making it possible to analyze numerous data and evolve artificial intelligence to learn on its own. Furthermore, it was possible to overcome that limitation with the development of a deep learning algorithm composed of a neural network structure that mimics the human brain.
In May 1997, IBM’s supercomputer Deep Blue received attention again when it defeated Gary Kasparov, the then world chess champion, after several challenges. And in February 2011, IBM’s Watson beat two contestants to win the Jeopardy quiz show, an American television broadcast program. The world was surprised that a machine had beaten a human, and accordingly. Interest in artificial intelligence development led by IBM and others has also increased significantly again.
Existing computers have brought significant changes to human society simply by doubling human computing capabilities. Moreover, this was only accomplished through human-written programs and involved humans. However, computers have created artificial intelligence capabilities that learn and make decisions on their own without human intervention. Intelligence has been known as a unique ability that only humans have. Of course, other species of animals also have some intelligence, but the degree is so weak that it is not enough to be called intelligent. However, it was surprising that computers were able to possess intelligence that was thought to be possessed only by humans. This was accomplished using a method called machine learning.
Machine learning focuses on creating algorithms that allow machines to learn certain signals and patterns from data without human help based on mathematical optimization and statistical analysis techniques, predict what will happen next based on that, and make appropriate decisions. The trend of artificial intelligence research relying on this machine learning methodology is particularly evident in June 2012, when Google and Andrew Ng used a diff-learning algorithm, a field of machine learning, to identify cat images in 74.8% of 10 million YouTube videos. A major leap forward was reached by successfully carrying out a project to enable identification with an accuracy of . Deep-learning was made possible using the concept of a neural network (Yongsoo Cho, 2015). With the rapid development of cloud computing environments and the support of big data, differential learning was implemented.
AI strategies of global companies
With this dramatic breakthrough, artificial intelligence entered a major turning point. Deep learning techniques that teach machines how to recognize images and sounds are producing concrete results as they are applied to various fields such as image recognition, voice recognition, and translation. Currently, major global companies view artificial intelligence as the biggest growth engine of the future. With the participation of Google, IBM, Microsoft, Apple, Facebook, Baidu, Alibaba, Samsung, etc., the areas of application of artificial intelligence are rapidly expanding to include improving medical technology, genetic analysis, new drug development, and financial transactions.
Google has been leading the artificial intelligence field through mergers and acquisitions of AI-related companies since 2001. In 2014, it acquired British AI company ‘DeepMind Technologies’ for $600 million (KRW 660 billion). DeepMind is the company that developed ‘AlphaGo,’ which aroused interest in artificial intelligence by winning a Go match against 9-dan Lee Sedol. Google has also been at the forefront of utilizing AI technology, including starting the development of self-driving cars in 2009. We are also ahead in the robotics field.
Google’s goal was to develop a true cyber helper for smartphone users that reads emails, understands all of the user’s movements, searches for what the user wants before even asking, and reports the results taking into account the user’s desired situation. As part of this, we are preparing services using artificial intelligence, such as AI assistant ‘Google Assistant’, Internet of Things (IoT) hub ‘Google Home’, and AI mobile messenger app ‘Allo’.
IBM developed ‘DeepBlue’ and ‘Watson’ to show real-life examples of artificial intelligence and is leading the field of artificial intelligence along with Google. IBM is focusing on creating a huge artificial intelligence ecosystem by incorporating it into each industry. The goal is to utilize Watson’s data search capabilities and natural language usage capabilities to enable developers to develop a variety of artificial intelligence services. IBM is particularly showing outstanding ability in applying Watson to the medical field. Watson is developing into a system that uses cancer patient data and various medical data to detect cancer and provide optimal treatment. It is said that Watson performs more accurately than famous doctors.
Additionally, they recently announced a self-driving shuttle bus using Watson IoT technology. This bus collects data from sensors mounted on the outside of the vehicle and is driven by AI using this data. In addition, daily conversation with passengers is possible, so when passengers tell them their destination, the vehicle automatically navigates to the destination and the optimal route. In this way, IBM wants to build a large artificial intelligence ecosystem by incorporating Watson into real life.
Microsoft is focusing on making Cortana the digital personal assistant that users rely on first. Cortana has natural language recognition and machine learning functions, and is developing into a service that goes beyond information search and helps users with what they want by utilizing big data from the search engine Bing. Microsoft believes that in the era of computing where all machines evolve based on artificial intelligence and understand human language, ‘dialogue’ will become a key means of communication between humans and objects, and summarizes this as the concept of ‘dialogue with platform’. there is.
Apple is focusing on expanding the ecosystem of ‘Siri’, a voice recognition information search service. AI Siri is linked with external apps to promote service scalability. In addition, Siri, which has only worked on iPhones and iPads, is said to be included in Mac OS, the operating system (OS) for PCs and laptops.
Facebook belatedly established an AI research center in 2013 and released ‘Chatbot,’ an AI chat platform. The goal was to create an AI-based ecosystem with 1 billion users.
Meanwhile, China is actively pursuing the AI industry based on its huge market and enormous capital power. Baidu is known to copy almost all of Google’s business models. Therefore, the field of artificial intelligence is no exception. Baidu recruited Andrew Ng, who led Google’s artificial intelligence field, and appointed him as head of the artificial intelligence research center. Baidu’s artificial intelligence lab has the goal of developing intelligence that can truly learn on its own without human intervention. It is expected that this artificial intelligence technology will improve Baidu’s business capabilities.
In this way, global companies have a common strategy to create and lead the artificial intelligence ecosystem. To this end, artificial intelligence software technology is commonly released as open source. This is to secure more developer allies by disclosing the technology and accelerate the evolution of the artificial intelligence ecosystem. MS open sourced AI Cortana and Skype translation voice recognition technology, and Google open sourced TensorFlow. TensorFlow was created for machine learning and deep learning.
Facebook also unveiled Big Sur, an open source hardware for AI technology. Big Sur is a server used to learn machine learning data, increasing data processing speed. China’s Baidu also released its self-developed artificial intelligence software as open source. WARP-CTC, an AI technology developed by Baidu’s artificial intelligence lab, was applied to the machine learning technology needed for computers to recognize human speech.
Spread of artificial intelligence
In this environment, many startup companies are also entering the artificial intelligence field. Accordingly, artificial intelligence has begun to be widely used, ranging from recommending products by reading the tastes of users of services such as content and shopping, to writing articles and materials, and autonomous driving. The Washington Post, a leading American daily newspaper, announced that an artificial intelligence robot called ‘Heliograf’ would write articles covering the Rio de Janeiro Olympics held in the summer of 2016.
AlphaGo’s success is also the work of the startup ‘DiffMind.’ DeepMind was evaluated as having raised the level of artificial intelligence to the next level through the game of Go, which requires human creativity to operate and handle numerous cases. AlphaGo, which has become the best in the world of baduk, is said to continue to evolve by challenging the game of Starcraft.
AlphaGo is an artificial intelligence for the game of Go, but behind it, the core technology is the process of the machine learning how to learn and win through massive amounts of big data, and the process of continuously strengthening knowledge through repetitive learning. Therefore, these core technologies can be applied and utilized in many other fields.
Financial work will also be handled by artificial intelligence. ‘Chatbot’ is a type of AI technology and is a computer program that carries out human-like conversations. In the past, chatbots used a simple pattern matching method to recognize only predefined keywords and output the entered response. Representative examples include tweetbot and article bots used by media companies to write short articles. However, chatbots based on artificial intelligence have evolved further, making it possible to understand the context and respond when questions or commands are asked in natural language. Additionally, as conversations accumulate, it learns on its own to increase accuracy. It is implemented as artificial intelligence that provides investment advice on behalf of people by automatically learning specialized information such as vast amounts of financial transaction information and economic indicators and reinforcing the learned knowledge.
Recently, it was reported that a startup called ‘8%’, a peer-to-peer (P2P) financial company, developed a chatbot ‘Aida’ (Electronic Newspaper, July 10, 2016). For example, “My income is 30 million won, and my current working period is 2 years. When you ask, “I want to borrow 10 million won, what is the interest rate and loan limit?”, Ada answers and provides a related URL link so that you can use P2P lending. Users can communicate as if they were talking on a messenger and find necessary information or use services without accessing a website or making a phone call. Deep learning, which was applied to AlphaGo, was used in Ada.
In particular, there is infinite potential in combining artificial intelligence with the Internet of Things to create services that are completely different from those of today. For example, when various healthcare devices are connected to an intelligent medical platform such as Watson, medical services ranging from simple collection or monitoring of health information to diagnosis and prescription of diseases can be implemented.
Of course, most commercialized AI is still limited to the service area. There is still a long way to go before large-scale businesses using AI are realized. Another problem that needs to be overcome is that AI still makes errors. A representative example is the recent death of the driver of a Tesla electric car when its autonomous driving function failed to recognize a trailer. If a high level of trust is not supported, there are limits to the spread of AI throughout the industry. Self-driving cars being promoted by companies such as Google are also restricted from entering public roads due to their inability to overcome reliability issues.
Meanwhile, the United States is ahead in the development and development of artificial intelligence, but China is catching up with great momentum. Following the Internet, China has set artificial intelligence as a national goal and is launching the 13th five-year China Brain project to develop human-device interaction, big data analysis and prediction, autonomous vehicles, and military and civilian robots. It was included in the plan (2016-2020) (Byung-un Kim, 2016).
Artificial intelligence is now expected to change the economy, society, and culture just as the Internet did. Not only does it change the way we communicate, but the culture itself changes. Artificial intelligence will be connected to all industrial sectors and change the industrial landscape. On the other hand, artificial intelligence will have a huge impact on employment by taking away human jobs. AI can have an impact that goes beyond the Internet or smartphones.
