AI Taiwan

The Plan launched an innovative training model that replaces training with actual operations and incorporates industrial, academic, and research capacity in innovative field verification to overturn traditional methods for cultivating talents. It closely connects companies’ actual difficulties and AI talents to expedite corporate development in innovative application services and AI transformation in the industry.

The Plan uncovers diverse demands in industries and proposes 21 AI solutions: Companies that utilized the “challenges from the industry and solutions from talents” system in 2018 included companies from diverse industries such as ICT, medical, e-commerce, human resources, security, and manufacturing industries. The Plan integrated 32 operators including Tung’s Taichung Metro Harbor Hospital, Taiwan Shin Kong Security, KKBOX, CyberLink, 1111 Job Bank, and Asia Neo Tech Industrial and provided a total of 53 AI transformation requirements. It also facilitated the development of solution provider teams formed by academic and research institutions, Taiwan AI Academy, and startups. A total of 43 teams proposed 64 solutions and produced 21 qualified solutions to satisfy diverse industrial requirements.

Solutions highlight the potential of AI:
AI precision lung tumor detection: iiNumbers Co., Ltd. (Startup) and Tung’s Taichung Metro Harbor Hospital cooperated in the “AI-assisted lung tumor diagnosis” project and used 3D imaging detection to significantly increase the efficiency of doctors’ diagnosis by 100-fold and increase the correct diagnosis rate to 99%. The project team later conducted research and development in biomarker technologies with Tung’s Taichung Metro Harbor Hospital, China Medical University, and medical imaging system provider Medincom Technology and launched the biomarker technology in the “AI X Smart Transportation Innovation and Applications Conference”. AI greatly reduced errors in security monitoring: UMBO CV Team and Shin Kong Security’s collaboration in the “AI Imaging Applications in Security Monitoring” project used image recognition technology to reduce more than 95% of errors in reports and improved security manpower performance. The project has been adopted into Shin Kong Security’s operations for outdoor field tests. AI perfectly duplicates cosmetics used by magazine models: Robust Team and Perfect Corp. collaboration in “AI cosmetics decoding” extracted the colors of eyeshadows from works of cosmetics on magazines and provided solutions with more than 87% similarities with actual colors. It is used in Perfect Corp’s related apps with 700 million downloads.

Collaboration with international companies for the AI development in Taiwan: The Plan partnered with Google and organized two simultaneous AI teachers’ camps in Northern, Central, and Southern Taiwan in 2018. The courses included introductory ML/DL courses, TensorFlow, ML issue framework establishment, ML model establishment, and AI product and service design workshops and provided training to 298 participants. Personnel that completed all training programs were awarded with training completion certificates from Google. They will be able to use Google’s MLCC course resources in their courses and quickly increase and expand AI capacity and AI education in Taiwan. In addition to Google, we also collaborated with Microsoft to cultivate AI seed teaching staff. We integrated Microsoft’s online education resources and physical courses to provide guidance and hands-on operations in actual cases. We trained 50 seed teaching staff who obtained Microsoft’s MPP certification as well as a total of 209 international certifications.

Conclusion:
The “Challenges from the Industry and Solutions from Talents” model in 2018 focused on “learning from doing” and revolutionized the talent cultivation model in traditional industries to closely connect companies’ actual difficulties and AI talents. The system differs from other AI technology contests by focusing on the output of AI solutions for the overall industry and the AI requirements of industries. In addition to training talented solution providers in establishing remarkable AI models, the contest also allowed them to learn a wide range of skills such as system integration, business negotiations, project management, and knowledge of the industry and cultivate the solution teams’ integration capabilities for providing proof of concept. It integrated knowledge of the industry and AI technologies and facilitated mutual learning to establish new models for “actual operations”, “interdisciplinary applications”, and “industrial applications”.

Date：2019-04-12
Source：Board of Science and Technology, Executive Yuan