i卡會員

As ITRI IEK indicates, the focus in the ICT industry this year is AI on earth, AI on edge and AI everywhere. In other words, AI is both on the cloud and on the edge (i.e. edge computing). This will start to have a significant influence on the technological roadmap, R&D and product design beginning in 2018. The IEK forecasts the global market for edge computing to grow from $8 billion in 2017 to $13.3 billion in 2022, at a CAGR of 10.7%.

According to the IEK, the initial stages of AI development was driven by cloud computing. However, more computing tasks will have to be transferred to edge devices or proximity network equipment due to restrictions on network bandwidth, communication latency and data security. The solutions to the emergence of edge computing proposed by the industry involve dedicated computing chips, sensor fusion, network architectures and communication equipment. We have summarized five edge devices suitable for edge computing architecture. 

There are a number of key issues on the horizon in 2018 in terms of the ICT infrastructure development. For example, cellular V2X (C-V2X) based on 5G is likely to be the standard for connected cars. The scope of blockchain applications will rapidly develop. Information security for industrial IoT will need to be further improved. Meanwhile, the potential problems in morality as a result of AI has prompted governments around the world to start the design of regulatory regimes. The prevalence of full-screen smartphones in the near future is expected to set a new tone for the development of components. Over the mid-term and in the long run, quantum computing is walking out of science fiction and into the commercial world. Below, IEK examines the key issues and relevant developments in the industry.

Key Issue No. 1: AI chips for edge computing are in the spotlight. The trend is for diversity and variety.

The first impact of edge computing is on the development of computing chips. IEK believes that edge computing chips will be specialized, for smart speakers, drones, robots, AR/VR, smart monitoring and other applications. In the smartphone market, Apple has launched A11 Bionic, and HiSilicon has introduced Kirin 970. Samsung and Mediatek are also planning to unveil a neural engine SoC (system-on-chip). In the autonomous driving market, Mobileye from Israel has developed EyeQ5, and NVidia has introduced Xavier SoC. Meanwhile, as accelerator chips have to optimize system specifications, hardware architecture and software algorithms, manufacturers have to come up with their own roadmaps and strategies. As global heavyweights are currently dominating the CPU and GPU chip markets for cloud computing, Taiwanese companies should focus on the development of AI chips for inference on edge devices. In this segment, FPGA/ASIC chips dedicated to image and vision should be a promising niche.  

Key Issue No. 2: Five edge devices as the first wave of AI edge computing

AI’s descent from the cloud to edge devices will see a plethora of AI-empowered edge devices. The devices for driverless cars, robots, monitoring, AR/VR and drones will be the first adopters of edge computing due to their requirements for low latency and high data transmission rates. Some Taiwanese companies have been involved in these five sub-verticals for a while and are well-positioned for the further development of hardware. Given the R&D resources and momentum in image processing and machine vision in Taiwan, manufacturers are advised to direct their efforts toward the development of vision-based edge devices.

Key Issue No. 3: AI edge computing drives the demand for new networking and computing equipment.

The global market for IoT devices is expected to grow from 8.4 billion units in 2017 to 20.4 billion units in 2020. By 2022, 75% of data processing will not be completed over the cloud, but on edge devices close to end users. 

Edge computing aims to reduce latency through a tiered architecture. International networking and server manufacturers are seeking to develop edge computing products by leveraging their existing product portfolios. As network nodes are likely to be at the forefront of edge computing, manufacturers from Taiwan may see opportunities and challenges in the supply chain of the telecommunications industry. The design focus of edge computing devices is on how to handle a distributed and heterogeneous computing environment, with key features such as high/low temperature tolerance, reliability and remote access. The integration with AI chips and the corresponding adjustments to system architecture are also critical. The server manufacturers in Taiwan have played a significant role in the supply chain of cloud data centers, and hence are equipped with the capability for AI chip integration and system architecture design. It is important, therefore, to enhance equipment reliability and strengthen service-level agreements (SLAs) in order to achieve the carrier-grade standard required by the supply chain. Networking equipment vendors may establish a strong foothold in edge computing by focusing on AI chip integration and corresponding system architecture design.

Key Issue No. 4: AI-empowered edge computing and sensor fusion create new types of human-machine interfaces.

The combination of AI and sensor fusion is expected to push international players and start-ups into the development of new technologies for human-machine interfaces (HMI). The solutions that integrate AI and sensor fusion will be ubiquitous, on wearables, robots and IoT platforms, in the form of AR/VR, gloves, skin, humanoid robots, intelligent machines (e.g. driverless cars, bicycles, drones, unmanned ships and handset robots), smart buildings, smart homes and smart infrastructure.

Taiwanese manufacturers have been shipping small volumes of sensors and algorithm solutions in the wearable market such as smart watches/bands, AR/VR and smart shoes. Currently, smart watches and consumer electronics still account for the bulk of this market. International semiconductor companies such as STM, ADI, Sony, TI, SiliconLabs and Omnivision, as well as new ventures, have been spending on the R&D of actuators and energy harvesters. In Taiwan, the bulk of R&D efforts are being made by academic and research institutes. Industry players have yet to get involved in this area. The sensor market for AI and sensor fusion products is forecast to show significant growth driven by the demand from a variety of applications for AI and sensor fusion over the next five years. The market for human-computer interaction (HCI) is expected to increase from $10 billion in 2017 to $23.8 billion in 2021, at a CAGR (compound annual growth rate) of 18.8%.

Key Issue No. 5: Information security is a growing headache for B2B IoT.

The increasing popularity of IoT applications and the ensuing complexity of the network environments have created vulnerability to different hacking techniques. Information security has become a serious issue for IoT. B2B industries such as manufacturing and medicare, where information security awareness is relatively weak, have become the target of hackers. IT security vendors are advised to develop and enhance protection over the networking environment (including gateways) in domains such as manufacturing and healthcare where robots are being deployed in high numbers. It is also necessary to integrate with next-generation firewalls in order to handle the new ITxOT paradigm due to the use of industrial robots controlled by computers. Meanwhile, vendors should assist robot hardware providers to improve their manufacturing process from the perspective of information security. In the long run, IT security vendors from Taiwan should enter the domains of behaviour analytics, accurate machine learning, real-time scanning and edge protection. It is also necessary to develop solutions for scenarios in different fields such as manufacturing and medicare. This will augment the information security for industrial IoT in Taiwan and meet the demand for industrial IoT security.  

Key Issue No. 6: Blockchain innovative applications gradually enter the mass market

The year 2017 witnessed the proof of concept for blockchain applications. In 2018, the single-point validation will accelerate into commercial deployments. The mushrooming of different blockchains without inter-chain connections will prevent data exchange and restrict the future of the technology. The focus going forward should be on the cross-chain protocols. Corporates, start-ups, academic and research organizations in Taiwan have made good progress in the development of blockchain technology and applications. In fact, the Taiwanese blockchain players have established knowhow and extended beyond the financial sector. However, the solo approach of individual blockchains in Taiwan is detrimental to the long-term prospects of their commercial potential, given the limited size of the domestic market. It is suggested that the industry should seek strategic alliances so as to develop cross-chain solutions. This will enhance the blockchain companies’ capability in working with various industries and create business opportunities in different sectors.

Key Issue No. 7: Cellular V2X marches on

According to statistics by Ovum, a research firm focusing on the global telecommunications market, there were a total of 50 5G trial projects around the world by the end of the third quarter of 2017. Approximately 16% of these pilot programs are related to cellular V2X, with tests run by telco operators such as NTT DoCoMo, Orange, Telenor, Korea Telecom and SK Telecom. It is evident that cellular V2X is one of the main focuses for 5G telecom services. In Taiwan, the networking and communications industry has not worked with auto makers very much. Along with its traditional focus on telecom clients, the industry can also prioritize the development of cellular V2X modules, devices, gateways and base stations by complying with the 3GPP standards in order to capture the market opportunities. As trial runs are a prerequisite before V2X service commercialization, field tests can be facilitated to assist the industry in the collection of relevant data such as image stitching, latency mitigation, reliability enhancement and technology/product improvements. These efforts will help companies in Taiwan to grasp business opportunities in the cellular V2X market.

Key Issue No. 8: Full-screen smartphones dictate product design and technology upgrades.

The year 2018 will see the adoption of full-screens not only by high-end but also by mid-end smartphones. As a full-screen aims to deliver a better user experience, the previously independent operations at the component level are being forced to integrate. For example, the embedding of fingerprint recognition into panel modules may be extended from “under-display integration” to “in-display integration” (i.e. as part of the panel manufacturing process). Meanwhile, the continued miniaturization will step up the level of difficulty in production and hence the costs. The change in appearance design will also require the re-design of components. Both the improvement of the production process and the new look of components will help to create new business opportunities and better pricing and demand/supply dynamics in the components market. Taiwanese companies have built a solid foundation in both display modules and sensor modules and can leverage this opportunity to develop new frontiers.    

Key Issue No. 9: First commercialization of the quantum computer marks the beginning of the next generation of computing.

The number of patents on quantum technology saw an annual growth of 12~39% in 2014-2016, indicative of significant breakthroughs. D-Wave Systems, founded in 1999, is the global leader in quantum computing. It is currently working with NASA and Google to explore algorithms for machine learning. Intel and IBM have produced quantum computing chips, and Microsoft has introduced Q#, the programming language for quantum computing. The collaboration between software and hardware heavyweights aims to bring quantum computing into commercialization within the next five years.

The era of quantum computing will surely lead to a new supply chain. New hardware and software are required. It is also necessary to use new materials such as superconductors and niobium (Nb), a precious metal. It has to operate in the extreme environment of zero Kelvin (minus 273 degrees Celsius). The access to new materials, the understanding of the material characteristics and the requirement for precision in manufacturing and reliability are all challenges ahead. They are also hurdles for the companies in Taiwan seeking to enter the quantum computing industry.

Key Issue No. 10: Governments start to intervene in the development AI on the ethical borderline.

The development of artificial intelligence comes in stages, from algorithms, perceptions, awareness, decision-making and creation. The focus in the perception stage is on pattern recognition. Decision-making is about the choices and executions from artificial intelligence. In 2016~2017, governments around the world established committees to devise AI ethics codes, and published regulations and principles for AI developers, vendors and manufacturers. The three pillars in the codes of AI ethics are the behavior of machines, human-machine relationships and the replacement of humans. The binding power of these guiding principles will affect the direction and profitability of products under development. For instance, the consensus between governments is that responsibilities should be clearly defined for any black box, so that machine power can be cut off if and when needed. It is expected that governments will come up with their own ethical codes, laws and regulations. The companies in Taiwan who wish to continue the development of robots, unmanned vehicles, drones, and driverless cars should stay on top of policy implications and respond appropriately with product designs.