- Artificial Intelligence Has Come Closer: Use of Artificial Intelligence April 02, 2021
-
by Professor Zoonky Lee, Graduate School of Information
Difference Between Humans and Animals
According to the Old Testament, humans built the Tower of Babel to reach heaven and tried to make their name known. Seeing this human arrogance, Jehovah confused the people’s language so that they would not be able to understand each other in order to prevent them from doing this in the future.
The development of mankind is in line with that of knowledge generation methods. The most important key to the knowledge generation is connecting knowledge with others through communication. Therefore, restricting communication slows down the process of knowledge generation. The most decisive reason why humans are at the top of the ecosystem and are able to create a civilization as it is now is its superior language ability compared to other homo species (e.g. Neanderthal, Australopithecus, Homo erectus, etc.). Unlike other animals that have an ability to communicate simply about food discovery and predator aggression, humans can exchange opinions on invisible abstract concepts (e.g. love, belief in God, state, currency, etc.) and create new knowledge on top of knowledge of others based on their own experience. Later, it developed into the ability to create letters, and as it was decidedly made into a book, humans’ ability to store, accumulate, distribute, disseminate their knowledge and create the new one has extensively developed.
Together with the invention of the computer, the digital revolution enabled humans to communicate with anyone on the planet through a tiny, always-carrying gadget called a smartphone. Furthermore, digital revolution created a new indicator of knowledge generation and accelerated its process by empowering an ability to easily find and disseminate knowledge accumulated by humans through such things as the Internet and search.
History of Artificial Intelligence
Early artificial intelligence (AI), along with the development of computers, aimed to implement a system that makes judgments like those of humans by transferring human knowledge accumulated thus far to computers. Such AI called the ‘Expert System’ led to the success of early AI by imitating judgments by doctors, lawyers, and factory process experts by inserting their knowledge.
However, this research, which has been conducted since the mid-1980s, rapidly cooled down in the 1990s since it was found that computers had limitations in expressing human knowledge. As an alternative, a new AI system called ‘Neural Network’ was proposed. It enabled reasoning by inserting data rather than using the existing method of putting human knowledge into a computer.
Artificial neural networks, despite its expandability and potential, faced difficulties in mathematical calculations from the mid-1990s to the early 2000s, and the number of related studies rapidly decreased. Around 2010, however, a new algorithm called ‘deep learning’ made a breakthrough and now, AI is enjoying a new golden era through ‘AlphaGo’ and ‘GPT-3,’ etc.
Knowledge Generation Fostered by AI
The reason why we should pay attention to deep learning-based AI is that the system itself has the ability to generate knowledge, while the existing AI attempts to imitate humans by simply mapping human knowledge. AI’s ability to generate knowledge is the one that shows new results by finding patterns that humans cannot find when data is entered. It was only five years ago when we watched AlphaGo’s baduk (Go) match and saw a world champion losing against AI. Since then, if you look at the current world of baduk, professional players are changing their method of practicing baduk towards using AI. In the traditional baduk, research has been conducted mainly on the knowledge exchange between humans and a record of accumulated knowledge in the past. On the other hand, AI is now showing moves that humans could not think of, and humans are creating a new way of creating knowledge by studying how to accept them.
Yet, modern AI works properly only in specific areas of a given domain (e.g. baduk (Go) where the rules and 19 by 19 grid boards are set or video consisting only of the color values of the grid). When it comes to an open space, its use is limited due to a lack of basic knowledge or common sense. Nevertheless, AI currently has a new judgment ability based on the enormous amount of data accumulated by humans.
Use of AI
Accordingly, human ability and competitiveness in the future lie in making better decision-making through new knowledge generation, using these shortcomings and advantages of AI. For example, if you watch a recent freestyle chess tournament (i.e. a chess tournament where you bring all kinds of computers you want and play as a team), you have an ability to read the opponent’s tactics, and the human expert, who uses an AI that can simulate through a laptop, is ahead of the supercomputer.
Such combination of human and AI can be already found in many cases. In terms of automobiles and furniture design, there is an ongoing attempt by experts, who understand the engineering architecture of the structure, to make AI design and humans select with an aesthetic sense by considering the constraints of the machine. During this process, AI provides various types of designs that existing humans could not imagine, and final design decisions can be made based on human judgment. Also, for image reading decisions by doctors, it can be convenient if AI can replace their role in simple readings. However, if AI makes a decision that is different from doctors or that they do not know of, they can learn new information by studying the ‘reasons why AI made such decision.’
Now Is the Time for Everyone to Learn How to Use AI
Humans, who are distinct from other animals with higher language, expanded their scope of knowledge generation by creating letters, books, and digital devices. Now we are at the entry of a new era of knowledge generation with the use of AI. To this end, an interaction between humans and AI is needed by going beyond simply using the results of AI.
The interaction requires an ability to simulate the results of AI by using one’s own expertise and an ability to interpret the results like the winner of a freestyle chess match. The important point here is that a professional chess player at the match is not an AI developer, but a ‘user.’ In other words, what most professional players would need in the future is not an ability to develop algorithms (of course, the AI development field will continue to increase in demand), but to understand and use AI.
In school education, there was mostly research on AI algorithms from computer science perspective at the beginning. In recent years, however, it is expanding to AI human-computer interaction (HCI), behaviors of AI (e.g. prejudice and errors of AI), AI, law and ethics study across the humanities, society, business, and medicine. Furthermore, education on the use of AI in each application field (e.g. medicine, investment, marketing, etc.) is becoming increasingly important. In addition to this trend, AI research has continuously introduced and enhanced automated machine learning systems such as AutoML in recent times, thereby creating an environment in which non-experts can easily apply AI. Given the current speed of AI development, it is expected that we will soon be at the crossroads between mere substitutes for and super professional in AI.
-
SINCHON·INT′l
SINCHON·INT′l
- MIRAE
show mobile menu
mobile menu
