The 'AI National Strategic Project' is also set to proceed after completing the formation of the project team in April.

A total of 163 billion won will be invested in artificial intelligence-related budgets this year, a 47% increase compared to last year.

The Ministry of Science, ICT and Future Planning announced that it will fully pursue R&D related to artificial intelligence technology this year, focusing on AI software (73.9 billion won), AI hardware (25.8 billion won), and basic technologies (63.3 billion won).

Specifically, AI software is being intensively expanded in core source technology fields such as language/visual intelligence, learning, and next-generation reasoning (40.7 billion won in 2016), and major projects include the AI National Strategy Project (23.9 billion won), Flagship Project (14.5 billion won), AI Robot Convergence (10 billion won), and other major AI R&D projects (25.5 billion won).

The 'AI National Strategic Project' advances the development of source technologies (Exobrain, DeepView, GenieTalk, etc.) in the fields of language, vision, and speech intelligence, which have high industrial demand, and conducts leading research on AI technologies such as next-generation learning inference to secure mid-to-long-term technological superiority. First, we decided to recruit a project director by the end of this month, form a project team by the second week of April, and then proceed with project planning and business.

Artificial intelligence hardware will continue to expand research in fields that serve as the foundation for AI development, such as intelligent semiconductors and supercomputing (19.2 billion KRW in 2016). Major projects include intelligent semiconductors (16.6 billion KRW), supercomputing (6.2 billion KRW), and neuromorphic chips (3 billion KRW).

Exobrain's Key Language Processing Technologies Scheduled for API Release in September

The plan for supercomputing technology development is to enable the smooth execution of high-capacity AI software through high-performance computing infrastructure, and efforts are also being made to develop core technologies for intelligent semiconductors and neuromorphic chips that optimize machine learning software execution at the chip device level.

Basic technology provides long-term expanded support (50.7 billion KRW in 2016) for fields such as brain cognition/brain engineering and industrial mathematics, which provide the theoretical foundation for artificial intelligence technology. Projects are being carried out in fields such as brain cognition/brain engineering (61.8 billion KRW) and industrial mathematics (1.5 billion KRW). In addition, the total budget for AI-related initiatives, including the expansion of application services (20 billion KRW) and the establishment of industrial infrastructure such as workforce training (42.7 billion KRW), amounts to 225.7 billion KRW (a 63% increase compared to last year).
These initiatives are in accordance with the 'Strategy for the Development of the Intelligent Information Industry' (March 2016) and the 'Mid-to-Long-Term Comprehensive Measures for an Intelligent Information Society' (December 2016), both announced last year. The Ministry of Science, ICT and Future Planning is also striving to ensure that the results of government AI R&D can be easily utilized in the private sector.

To this end, key language processing element technologies of Exobrain, which is being developed as a core technology for natural language analysis and question answering, will be released in API form this coming September so that industry, academia, and research institutes can freely utilize them. The seven types of Korean analysis APIs to be released this year include morphology, named entity recognition, polysemous word analysis, dependency syntax analysis, semantic role recognition, lexical similarity analysis, and homonym analysis. Exobrain, which will proceed with commercialization technology development in the legal, patent, and financial fields until January 2020, is planned to be developed into a professional question-and-answer system at a commercialization level.

DeepView, which has been developing since 2014 as a source technology for visual intelligence that recognizes and understands objects (40 nouns) and actions (15 verbs) in photos and videos, also discloses visual knowledge representation technology that expresses the content of videos as knowledge and data for visual intelligence training.

An official from the Ministry of Science, ICT and Future Planning stated, “We will participate in a global competition (ImageNet) to verify our technological capabilities and pursue the application of visual intelligence technology to real-world problems for industrialization,” adding, “We plan to verify performance by applying large-scale CCTV content analysis technology to urban video and other environments.”