The Hankook Ilbo: The Balance of Seeing the World

The competition for super-giant AI technology, which is called ‘the future of artificial intelligence (AI)’, is hot. Super-giant AI with abilities that are one step closer to humans has infinite potential to be used in all fields of the future industry. So, if a company is preparing for future food, it is a situation in which it must not be left behind or fall behind in the super-giant AI competition that will unfold soon. According to the AI ​​industry on the 30th, global big tech companies such as Google and Microsoft (MS) are jumping into the development of super-giant AI early, while domestic conglomerates and information and communication (IT) companies are moving faster. Super-giant AI is an AI that thinks and judges like a human by learning massive amounts of data on its own. The biggest reason why super-giant AI is attracting attention is because of its wide range of applications. Experts predict that, just as the Internet changed the world in the past, super-giant AI will change the daily life of human beings in ways they could not have imagined. In fact, super-giant AI is highly likely to replace humans not only in business-to-consumer transactions (B2C) such as customer centers, but also in all industrial processes such as product development, product design, design, and distribution. Large corporations that provide a variety of services through affiliates cannot but covet a general-purpose AI that can be used in any field if developed once. The level of AI, which was only considered as an AI speaker’s voice assistant, has recently been advanced enough to be used in medical research. According to the international scientific journal Nature, Google’s AI developer DeepMind has reached the stage of developing new drugs by researching protein amino acids using AI. Microsoft and Nvidia worked with AstraZeneca to build AI for drug discovery. Just as AI makes sentences with conversation data learned in advance, AI who has learned molecular structure and clinical data suggests various ways to develop new drugs. The most advanced in this fierce technology competition is Google (Deep Mind), which developed the Go AI ‘AlphaGo’ and defeated the highest number of humans one after another. Google unveiled a super-giant AI switch transformer with a scale of up to 1.6 trillion parameters (parameters) in February. As the more parameters, the more sophisticated the AI ​​inferred results, and each company is fighting over who gets the bigger parameters. Microsoft also introduced ‘MT-NLG’, a language model with 530 billion parameters in October last year, and DeepMind introduced Gopher with 280 billion parameters in December last year. Among domestic companies, Naver is a representative company that is speeding up the development of super giant AI. Hyper CLOVA, unveiled by Naver in May last year, is the first super-giant AI based in Korean. Naver built AI using its vast data, such as blogs and news, and its performance is world-class. With a scale of 204 billion parameters, it has better capabilities than GPT-3 (175 billion parameters), which is the standard for super-giant AI. Kakao also unveiled KoGPT, an AI language model specialized for Korean in November last year. This is the Korean version of GPT-3. LG also introduced the super-giant AI EXAONE with 300 billion parameters in December last year. This is the achievement achieved within a year of investing in full-fledged investment by putting AI into the future and putting company-wide capabilities into it. A united front between academia, research institutes and large corporations was also established. SK installed a task force under SK Telecom in May of last year and is in the midst of developing a super-giant AI based on GPT-3 together with the National Institute of the Korean Language. Recently, AI assistant A. (Adot) was also announced. KT is developing a super-giant AI with 200 billion parameters in cooperation with the Electronics and Telecommunications Research Institute (ETRI) and the Korea Advanced Institute of Science and Technology (KAIST). But there are many hurdles to overcome. First, there is a lack of large-scale data to train AI. Although the basic code is shared as open source, it is the quality and quantity of data that determines the intelligence of AI. As domestic AI research and infrastructure is one step behind that of overseas, some point out that government support is urgent to build super-giant AI. Data-related regulations need to be revised in order to prevent the recurrence of the Iruda incident, which caused controversy over personal information infringement last year. Jang Byung-tak, head of the AI ​​Research Center at Seoul National University, said, “Just as Google develops services by taking all data from around the world, the super-giant AI field is likely to eventually focus on several platforms. It is open,” he said. The manpower shortage is also a long-awaited challenge. Sang-Kyun Cha, dean of the Graduate School of Data Science at Seoul National University, said, “Data science is important to create a super-giant AI, but there is no professor to teach it. “He pointed out.

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.