07.2020 Cover Story
Infinite applications, FET brings AI into daily life
Far Eastern Magazine / Editorial Room
With the progress of science and technology, AI (Artificial Intelligence) has already quietly entered our lives and been developed in various fields. As a digital pioneer of the Far Eastern Group, FET has taken an early step in the layout of AI technology with achievements in textile fiber, cement, steel and other industries, including: AI intelligent streetlights, cloud intelligent environmental control / environmental safety, AI information security management platform, AI face recognition and other applications, leading individuals, enterprises, and even FET itself into a new era.
Everyone has feeling of intelligent life
Artificial intelligence, like electricity and the Internet in the new era, can be used by all walks of life. The AI applications developed by FET is not only provided to individual users, but also to enterprise customers and used to optimize internal operations.
In personal applications, the e-commerce customer experience is enhanced through artificial intelligence, and the transaction rate is improved. For example, friDay searches for best-selling products through data, and with data such as search volume and product performance within the platform, as well as external Google trend, hot sales ranking of the competitors, price comparison website, PTT and other data, it forecasts hot selling products and provides suggestions for individual users to select products.
In addition, FET is also committed to the development of smart home. Among them, smart speaker is one of the indispensable devices in smart home AI application, especially the voice part, which is also an important stage of artificial intelligence. It is necessary to make the smart speaker understand the voice and tone of all kinds of consumers, and it should be suitable for all ages. Therefore, before the official launch of the smart speaker, FET first collected thousands of voice recordings, and through artificial intelligence analysis, it enabled the smart speaker to "understand" various utterances and instructions. In 2018, FET took the lead in Taiwan by launching the “Little Fox” smart speaker. And it was favored by the parents and the elders as “desk-mounted voice assistant" and became a leading brand of Taiwan's localized smart speaker. In 2019, FET launched two portable smart speakers. In addition to continuously expanding digital content services and strengthening the "voice value-added service ecosystem", in September, a comprehensive layout of the "smart control ecosystem" was announced. It successively launched smart home networking devices and expanded the connection with home appliance brands. With the smart speaker, the Internet of smart home was integrated and the development of intelligent control territory was accelerated.
In the future, FET will continue to expand the product line of smart speakers to provide consumers with more services and value-added content. At the same time, it will actively use the speakers to drive a variety of home-based IOT appliances, so that consumers can enjoy the service of modern technology at home.
Intelligent prediction, helper of all walks of life
AI is also widely used in enterprises. Far EasTone Telecommunication is committed to developing Internet of things (IOT) applications in recent years. Taking the smart street lamp launched in 2019 as an example, it can establish the recommended value of street lamp illumination control according to the flow of people, climate and environment, so as to meet the dual needs of lighting and energy saving. At the same time, based on device information, property and maintenance behavior, a prediction model is established to make the scheduling strategy of street lamp maintenance more economical and efficient. Another energy management system is to use the internal and external climate and the past electricity consumption behavior to establish the indoor temperature prediction model and power consumption prediction model, and to establish the optimal equation of electricity consumption, so as to achieve the goal of power saving of central air conditioning system.
In terms of operation, AI is also a good helper for increasing revenue and reducing operating costs and risks. Taking the most troublesome debt problem in the telecom industry as an example, FET predicts the debt risk through AI and makes a more flexible design of bill collection consumption quota. With the expansion of the demand for text services, it also uses Robot to replace the simple type of service volume to improve customer service efficiency and reduce costs. In addition, FET also develops its own text classification technology and excellent data tags, and builds ML (machine learning) algorithm model and crawlers to solve the problem that a large number of websites cannot be classified, optimizes the old system on a large scale, and makes marketing and sales more accurate.
The application of FET artificial intelligence is also extended to Far Eastern Group affiliated enterprises. AI intelligent fabric inspection machine and AI intelligent printing machine cooperated with FENC can identify the types and characteristics of good products with high-quality images through visual AI. And with intelligent learning, it can improve the efficiency of fabric inspection at the time of shipment and reduce the customer complaints caused by defective products.
AI can not only help enterprises upgrade, but also fight crime. In 2018, FET was informed by the Criminal Police Bureau that "the signal of FET in Kinmen is too good", which has become the first choice of fraudsters. Many of the fraud calls using IF cards come from Xiamen, which is adjacent to Kinmen. In order to solve the problem from the source, an engineering team was sent to Kinmen to adjust the base station signal, and took a boat to the sea center to test the signal strength, so as to ensure that users of Kinmen would not be affected, which made fraud calls reduce by nearly 40%. Since information is the basis of machine learning, FET then carried out big data analysis on the information of fraud calls in the past. features like daily call volume, dialing behavior and average call time were effectively separated, and data scientists selected appropriate AI algorithms and tried various model parameters to establish machine learning models, and finally find out 36 key clues of fraudsters. Last year, the accuracy rate of fraud detection through the “automatic detection system of fraud prediction” reached 92.5%, which also enabled FET become the first telecom operator in Taiwan to detect fraudulent calls through AI machine learning and obtain the patent of the Ministry of economic affairs.
Not only can AI be used to establish prediction models for human behavior, FET and Water Resources Agency jointly published the first "online collapse and turbidity prediction system" in Taiwan, which AI is used to predict the natural environment. However, the objects, data and model types used are different. In short, machine learning is divided into two types: regression and classification. The former is usually used to predict a number, including: house prices, closing points of the stock market tomorrow, weather conditions in the future. For example, the actual price of a product is 500 yuan, and the predicted value can be 499 yuan through regression analysis. In addition, the collapse and turbidity prediction system is also a "regression" problem. A number is obtained by the model to predict the probability of collapse and the rise of turbidity after 6 hours, so as to strive for valuable contingency time for disaster prevention. As for AI's detection of fraud, it belongs to the "classification" problem, which means that the model will judge "yes" or "no" by randomly sampling a person.
Intelligent challenges, infinite business opportunities
AI can provide a wide range of solutions for different application scenarios, which is far more than people think. Although the future development is unlimited, it is inevitable to face the following challenges.
1. Business opportunities and demands: due to the difficulty in measuring the effectiveness of artificial intelligence, the high cost and manpower input, and the difficulty in integrating the existing company's system, the current demand is not clear. It is suggested that companies should clarify the purpose before putting artificial intelligence into practice. For example, what are the problems that they hope to solve through artificial intelligence? How to apply AI to products, services and processes in the future? Only by finding out problems and establishing requirements can we develop truly effective AI applications.
2. Data quality: artificial intelligence must be built on the right data and quality, and the data can be obtained continuously. However, most companies have limited data available, or it is difficult to integrate different types of data such as text, image, picture, voice, etc., which affects the application development and effectiveness of artificial intelligence.
3. Talents and tools: at present, there are limited AI talents in the market. However, cultivating talents requires experience, time and resources, and most enterprises need business opportunities to be willing to invest. Moreover, there is still a lack of integrated open source application development tools, which makes the development process slow and easy to have information and process security concerns.
4. Potential risks: the rise of artificial intelligence applications has brought convenience to life. However, the liability attribution (e.g. self-driving accident) and information security in case of accidents also lead to the legal liability risk of service providers. In addition, the application of artificial intelligence will inevitably replace part of the original manual handling, how to achieve a balance between technological advancement and the impact of the employment market is also a big problem.
However, with the evolution of communication technology, the coming 5G, with its characteristics of large bandwidth, wide coverage and low latency, will contribute to the realization of edge computing and further advance the development of artificial intelligence. In particular, it is expected that there will be promising achievements in transportation and UAVs in the future. With the combination of artificial intelligence and 5G, intelligent transportation can really enter into V2X (vehicle to everything) intelligent collaboration, strengthen the vehicle's perception of the environment and detection range, improve road traffic safety and reduce casualties. As for UAV, the HD video streaming obtained after the installation of monitoring can be used in environmental monitoring, accident risk and disaster prevention investigation. As in Japan, KDDI telecom has used UAVs to provide safe, rapid and cost-effective infrastructure monitoring services.
Far EasTone Telecommunication has been transforming from a traditional telecommunication industry which only provides voice and text message to a service provider of data and digital innovative applications. In the intelligent era full of multiple business opportunities, FET has injected fresh water into the telecom industry. It is believed that the goal of various AI applications will come true in the near future. Wei